---
schema_version: "newruntime-agent-readable-v0.2"
type: "project"
stable_id: "project:github-projects-for-ai-engineers"
slug: "github-projects-for-ai-engineers"
title: "GitHub Projects for AI Engineers"
description: "A Telegram-corpus-derived GitHub radar plus a source-backed map of the models, orchestration, retrieval, memory, security, automation, and infrastructure layers around production AI systems."
retrieval_nugget: "A Telegram-corpus-derived GitHub radar plus a source-backed map of the models, orchestration, retrieval, memory, security, automation, and infrastructure layers around production AI systems. Turn the Telegram export and two public ecosystem references into a ranked repository radar and a navigable map of the wider production AI stack."
status: "published"
project_status: "candidate"
one_line_outcome: "Turn the Telegram export and two public ecosystem references into a ranked repository radar and a navigable map of the wider production AI stack."
wow_moment: "One HTML or JSON route now exposes 140 ranked repositories plus 98 official tool links across 13 capability lanes and nine production-depth levels."
buildability_class: "B2"
bottleneck: "GitHub metrics drift quickly; every refresh needs a current GitHub read and a relevance pass against the corpus."
time_to_first_result: "One local content pass plus a GitHub metric refresh"
estimated_cost: "Free, except the curation time and GitHub metric refresh"
code_level: "AI engineer, developer advocate, or technical editor"
data_touched: "Public GitHub repository pages and local Telegram export link roots"
write_actions: "Public site content only after owner-approved deployment"
handoff: {"inputs":["Local Telegram export repository-link roots","Public GitHub repository URLs","Current GitHub stars, forks, archived state, and canonical repository names","New Runtime topic taxonomy and project theme lanes"],"tools":["Local data extraction","GitHub public web/API reads","Repository normalization script","Astro content validation"],"build_steps":["Extract GitHub repository roots from the approved local export or existing project dataset.","Resolve redirects and drop placeholders, private repositories, duplicates, and non-AI-engineering projects.","Refresh public GitHub metrics and canonical full names for the surviving repositories.","Assign each repository to one theme lane and explain why it matters to New Runtime.","Rebuild the project JSON/Markdown/HTML routes and verify the ranked ledger still matches the refreshed dataset."],"acceptance_checks":["Every retained repository has a public GitHub URL, canonical full name, stars, forks, theme, engineering area, and reason.","The ranking is deterministic: stars descending, forks as the secondary signal.","Private Telegram provenance is not exposed; public citations point to GitHub repository URLs only.","The HTML radar and JSON dataset agree on project count, theme counts, and top-ranked repositories.","The ecosystem section and JSON dataset agree on tool count, capability lanes, production-depth levels, link kinds, and source attribution.","Astro build and agent index contract tests pass after the refresh."],"non_goals":["Do not publish the raw Telegram export or private discovery provenance.","Do not clone or execute the listed repositories during the first refresh pass.","Do not add repositories only because they are popular if they do not support AI engineering work."],"questions_before_external_access":["Which export snapshot or dataset is the source of truth for this refresh?","Is a live GitHub metric refresh approved for this run?","Should archived, redirected, or sponsor-only repositories be kept, dropped, or marked?"],"permissions":[{"purpose":"Read public repository metadata from GitHub.","access":"read","minimum_scope":"Public repository pages or unauthenticated GitHub API reads where possible.","confirmation":"Human approves a current metric refresh before network calls.","cleanup":"Delete temporary metric caches after committing the normalized public dataset, unless retained as a dated local artifact."},{"purpose":"Update public site content after review.","access":"publish","minimum_scope":"Astro content/data files for this project and only after owner-approved deploy.","confirmation":"Human approves publication/deploy separately from the local refresh.","cleanup":"Rollback the focused site release if the public radar exposes private provenance or incorrect repository facts."}],"reproducibility_notes":["Record the metric refresh date and the exact sorting rule in the public method section.","Keep dropped-repository reasons in a private local audit artifact, not in public source lists."]}
proof: {"verification_status":"locally-reproduced","fixture_outputs":[{"label":"Structured radar dataset","artifact":"dist/projects/github-projects-for-ai-engineers.json","route":"https://newruntime.com/projects/github-projects-for-ai-engineers.json","check":"The JSON route exposes the ranked repository dataset, proof contract, handoff routes, and public GitHub source URLs."},{"label":"Human radar page","artifact":"dist/projects/github-projects-for-ai-engineers/index.html","route":"https://newruntime.com/projects/github-projects-for-ai-engineers/","check":"The HTML radar renders the same project count, theme lanes, and machine-readable handoff links as the dataset."},{"label":"Agent build brief","artifact":"dist/projects/github-projects-for-ai-engineers/build-brief.md","route":"https://newruntime.com/projects/github-projects-for-ai-engineers/build-brief.md","check":"The build brief names the data source, GitHub metric refresh boundary, privacy constraints, and acceptance checks."},{"label":"Clean-room replay fixture","artifact":"scripts/project-proof-depth-v2-fixture-test.mjs","check":"The fixture replays the bounded deterministic-ranking task in a temporary directory, proves the initial failure of the wrong sort order, proves the final pass of the stars-then-forks rule, and verifies that an unrelated file is unchanged."},{"label":"Independent replay kit","artifact":"scripts/project-proof-independent-replay-kit.mjs","check":"The kit starts with a ranking function that violates the published sort rule, exposes no ready solution, protects every out-of-scope file by checksum, and refuses to emit a receipt without a non-author attestation and a passing final check."},{"label":"Independent agent replay receipt","artifact":"src/data/project-proof-receipts/github-projects-for-ai-engineers-independent-agent-v1.json","route":"https://newruntime.com/projects/github-projects-for-ai-engineers.json","check":"A separately spawned agent received only the isolated kit, observed the failing deterministic-ranking check, changed only src/rank.mjs, passed the final check, and produced an attestation that the primary verifier accepted."}],"repro_commands":["npm run project-proof:replay:test","npm run project-proof:replay:prepare -- --project=github-projects-for-ai-engineers","npm run validate:content","npm run build","npm run validate:agent-indexes"],"expected_artifacts":["A normalized repository dataset with canonical repo name, URL, stars, forks, theme, area, export hits, and reason.","A deterministic ranking pass sorted by stars descending, then forks as the secondary signal.","A normalized ecosystem dataset with official links, capability categories, production-depth levels, and explicit source notes.","A private dropped-repository audit when refreshing from a local export snapshot."],"provider_scopes":[{"provider":"GitHub","access":"read","minimum_scope":"Public repository metadata and repository pages only.","why_needed":"Stars, forks, archived state, redirects, and canonical full names drift over time.","approval_gate":"Human approves a live GitHub refresh before network calls."},{"provider":"Local Telegram export","access":"read","minimum_scope":"Repository-link roots and duplicate counts from an approved local export snapshot.","why_needed":"The radar starts from private discovery provenance but publishes only public GitHub repository URLs.","approval_gate":"Human names the export snapshot or existing dataset before extraction."},{"provider":"Public site","access":"publish","minimum_scope":"Project content, project dataset, and generated static routes for this radar only.","why_needed":"The public radar must expose the refreshed structured dataset and human visualization together.","approval_gate":"Owner approves commit, push, and deploy separately from the metric refresh."}],"known_limits":["GitHub stars and forks are time-sensitive; the public method must keep the refresh date visible.","The first pass does not clone, run, or security-audit the listed repositories.","Export-hit counts are discovery signals, not public provenance; raw Telegram data remains private.","The ecosystem map is a navigation and architecture aid, not a compatibility matrix or recommendation to adopt every tool.","The bounded replay proves the deterministic ranking rule on a small fixture dataset, not a full 140-repository refresh against live GitHub metrics.","The independent replay proves the bounded public workflow contract, not integration into a real external or production repository.","The independent reviewer identity is a local agent attestation and is not cryptographically verified."],"buildability_breakdown":{"assessed_at":"2026-08-15","assessment_level":"non-author-replayed","axes":[{"axis":"time","level":"high","evidence":"The clean-room ranking task reaches a checked result in one bounded local run, while the public blueprint targets one local content pass plus a GitHub metric refresh.","bottleneck":"A full radar refresh still depends on fetching current GitHub metrics for all 140 repositories."},{"axis":"code_burden","level":"medium","evidence":"The replay needs one focused comparator edit plus the existing test, and the public blueprint needs normalization scripts and content validation rather than a bespoke application.","bottleneck":"Curating theme lanes and relevance still needs an editor who knows the AI-engineering landscape."},{"axis":"integration_burden","level":"medium","evidence":"The replay uses only repository files, Node, and a test runner, while the wider blueprint touches public GitHub reads and Astro content routes that already exist in this repository.","bottleneck":"GitHub metric drift forces every refresh through an approved external read before the ledger is current."},{"axis":"operational_burden","level":"high","evidence":"The first proof is local and temporary, requires no server, and touches no shared or production state.","bottleneck":"Keeping the published radar current requires a recurring refresh protocol rather than one-off runs."},{"axis":"permission_clarity","level":"high","evidence":"GitHub read, local Telegram-export read, and public-site publish scopes are named separately with explicit approval gates in the build brief and acceptance contract.","bottleneck":"A live metric refresh and a public deploy still need two separate owner approvals."},{"axis":"reproducibility","level":"high","evidence":"A separately spawned agent received only the isolated public kit, reproduced the failing deterministic-ranking check, changed the one allowed comparator file, and passed the same acceptance check without inspecting the author solution.","bottleneck":"The bounded fixture proves the published sort rule, not a full 140-repository refresh against live GitHub data."},{"axis":"failure_recovery","level":"high","evidence":"The replay records the initial failing check, preserves an unrelated file byte-for-byte, performs no remote action, and leaves explicit residual-risk and approval sections.","bottleneck":"Rolling back a bad public radar release is outside this local proof."}]},"replay_evidence":[{"id":"clean-room-bounded-edit-v1","replayed_at":"2026-08-15","replay_kind":"automated-clean-room","instructions_source":"public-build-brief-and-acceptance","environment":"Temporary local directory with a bounded Node fixture and no network or remote writes.","outcome":"passed","commands":["npm run project-proof:replay:test"],"artifacts":[{"label":"Task brief","artifact":"task-brief.json","check":"Names the allowed comparator file, acceptance command, non-goals, and approval-gated actions."},{"label":"Focused implementation","artifact":"src/rank.mjs","check":"Changes only the named comparator file while the unrelated note retains its original checksum."},{"label":"Real test","artifact":"test/rank.test.mjs","check":"Fails while the initial comparator sorts ascending without a forks tiebreaker, and passes after the focused edit applies stars descending with forks as the secondary signal."},{"label":"Structured handoff","artifact":"handoff.json","check":"Separates implemented files, verification evidence, residual risk, and approval-gated actions."}],"limitations":["This replay proves the deterministic ranking contract on a bounded fixture dataset, not a real external repository integration or a live GitHub metric refresh.","No human or separately operated agent independently interpreted the brief in this clean-room replay."]},{"id":"independent-claude-replay-v1","replayed_at":"2026-08-15","replay_kind":"independent-agent","instructions_source":"public-build-brief-and-acceptance","environment":"Isolated temporary replay kit; the executor was a separately spawned agent session that received only the kit path, no parent conversation or main-repository context, and used no network or remote writes.","outcome":"passed","commands":["node --test test/rank.test.mjs"],"artifacts":[{"label":"Independent attestation","artifact":"reviewer-attestation.json","check":"Names an independent agent, records initial failure and final pass, lists only src/rank.mjs as changed, and attests that the author solution was not inspected."},{"label":"Verified replay receipt","artifact":"src/data/project-proof-receipts/github-projects-for-ai-engineers-independent-agent-v1.json","check":"The primary verifier checked protected-file hashes, rejected unexpected files, reran the acceptance test, and recorded zero unexpected files.","route":"https://newruntime.com/projects/github-projects-for-ai-engineers.json"}],"limitations":["This proves a bounded non-author replay of the public workflow contract, not integration into a production repository.","Reviewer identity is a local agent attestation and is not cryptographically verified."]}]}
last_verified: "2026-08-15"
record_date: "2026-08-15"
date_kind: "last_verified"
related_shift: "shared-context-infrastructure"
related_posts: ["ai-engineering-under-the-hood","cognee-agent-memory-platform","claude-code-long-running-loops","agent-api-design"]
topics: ["github","ai-engineering","agent-runtime","context-engineering","llm-ui","evals","model-infrastructure","agent-security"]
source_urls: ["https://github.com/microsoft/markitdown","https://github.com/firecrawl/firecrawl","https://github.com/langgenius/dify","https://github.com/langchain-ai/langchain","https://github.com/anthropics/claude-code","https://github.com/Shubhamsaboo/awesome-llm-apps","https://github.com/github/spec-kit","https://github.com/browser-use/browser-use","https://github.com/google-gemini/gemini-cli","https://github.com/openai/codex","https://github.com/punkpeye/awesome-mcp-servers","https://github.com/modelcontextprotocol/servers","https://github.com/vllm-project/vllm","https://github.com/openai/openai-cookbook","https://github.com/unslothai/unsloth","https://github.com/warpdotdev/warp","https://github.com/microsoft/autogen","https://github.com/crewAIInc/crewAI","https://github.com/aaif-goose/goose","https://github.com/agno-agi/agno","https://github.com/langchain-ai/langgraph","https://github.com/topoteretes/cognee","https://github.com/getzep/graphiti","https://github.com/openai/openai-agents-python","https://github.com/yamadashy/repomix","https://github.com/PrefectHQ/fastmcp","https://github.com/langchain-ai/deepagents","https://github.com/mastra-ai/mastra","https://github.com/deepset-ai/haystack","https://github.com/a2aproject/A2A","https://github.com/openai/evals","https://github.com/camel-ai/camel","https://github.com/confident-ai/deepeval","https://github.com/vercel-labs/json-render","https://github.com/dottxt-ai/outlines","https://github.com/tambo-ai/tambo","https://github.com/LMCache/LMCache","https://github.com/TencentCloud/CubeSandbox","https://github.com/0x4m4/hexstrike-ai","https://github.com/microsoft/magentic-ui","https://github.com/openclaw/openclaw","https://github.com/obra/superpowers","https://github.com/NousResearch/hermes-agent","https://github.com/x1xhlol/system-prompts-and-models-of-ai-tools","https://github.com/garrytan/gstack","https://github.com/bytedance/deer-flow","https://github.com/cline/cline","https://github.com/docling-project/docling","https://github.com/anthropics/claude-cookbooks","https://github.com/google/langextract","https://github.com/heygen-com/hyperframes","https://github.com/github/awesome-copilot","https://github.com/anthropics/claude-plugins-official","https://github.com/opendataloader-project/opendataloader-pdf","https://github.com/Fosowl/agenticSeek","https://github.com/google-labs-code/design.md","https://github.com/openai/symphony","https://github.com/agentskills/agentskills","https://github.com/allenai/olmocr","https://github.com/microsoft/agent-lightning","https://github.com/MemoriLabs/Memori","https://github.com/eigent-ai/eigent","https://github.com/NevaMind-AI/memU","https://github.com/NVIDIA/SkillSpector","https://github.com/langchain-ai/openwiki","https://github.com/cocoindex-io/cocoindex","https://github.com/MoonshotAI/kimi-cli","https://github.com/github/copilot-sdk","https://github.com/idosal/git-mcp","https://github.com/google-labs-code/stitch-skills","https://github.com/evidentlyai/evidently","https://github.com/GoogleCloudPlatform/agent-starter-pack","https://github.com/airweave-ai/airweave","https://github.com/steipete/agent-scripts","https://github.com/1weiho/open-slide","https://github.com/MinishLab/semble","https://github.com/google/agents-cli","https://github.com/mistralai/mistral-vibe","https://github.com/mixedbread-ai/mgrep","https://github.com/rivet-dev/agentos","https://github.com/google-gemini/gemini-skills","https://github.com/memodb-io/Acontext","https://github.com/openai/openai-agents-js","https://github.com/oceanbase/seekdb","https://github.com/openai/openai-apps-sdk-examples","https://github.com/benchflow-ai/skillsbench","https://github.com/chunkhound/chunkhound","https://github.com/langchain-ai/openevals","https://github.com/microsoft/work-iq","https://github.com/CopilotKit/generative-ui","https://github.com/multica-ai/andrej-karpathy-skills","https://github.com/mattpocock/skills","https://github.com/karpathy/autoresearch","https://github.com/DietrichGebert/ponytail","https://github.com/colbymchenry/codegraph","https://github.com/Panniantong/Agent-Reach","https://github.com/mvanhorn/last30days-skill","https://github.com/chenglou/pretext","https://github.com/pbakaus/impeccable","https://github.com/abhigyanpatwari/GitNexus","https://github.com/vercel-labs/agent-browser","https://github.com/tinyhumansai/openhuman","https://github.com/anthropics/financial-services","https://github.com/Yeachan-Heo/oh-my-codex","https://github.com/lightpanda-io/browser","https://github.com/supermemoryai/supermemory","https://github.com/vercel-labs/skills","https://github.com/rohitg00/agentmemory","https://github.com/openai/skills","https://github.com/ogulcancelik/herdr","https://github.com/mksglu/context-mode","https://github.com/pydantic/pydantic-ai","https://github.com/baidu/Unlimited-OCR","https://github.com/rowboatlabs/rowboat","https://github.com/google/skills","https://github.com/StarTrail-org/LEANN","https://github.com/InsForge/InsForge","https://github.com/zilliztech/claude-context","https://github.com/shadcn/improve","https://github.com/pydantic/monty","https://github.com/teamchong/pxpipe","https://github.com/max-sixty/worktrunk","https://github.com/agent-infra/sandbox","https://github.com/vercel-labs/zerolang","https://github.com/jgraph/drawio-mcp","https://github.com/cursor/cookbook","https://github.com/BuilderIO/skills","https://github.com/simonw/files-to-prompt","https://github.com/Agent-Field/agentfield","https://github.com/cursor/plugins","https://github.com/coinbase/agentkit","https://github.com/AI-Builder-Club/skills","https://github.com/SafeAI-Lab-X/ClawKeeper","https://github.com/openai/realtime-voice-component","https://github.com/langchain-ai/agent-protocol","https://github.com/senlindesign/taste-skill","https://github.com/zapier/AutomationBench","https://github.com/OpenDataBox/MemoryData","https://github.com/cisco-foundation-ai/fully-automated-prompt-optimization","https://github.com/AgentToolkit/altk-evolve","https://www.linkedin.com/posts/rathanuday_the-open-source-ai-ecosystem-is-evolving-activity-7398656721302249472-2C2R","https://www.bhuviai.com/agentic-ai-rag-tech-stack-layered-blueprint"]
dataset: {"type":"github_project_radar","ranking":"stars_desc_forks_desc","verified_at":"2026-07-23","themes":[{"id":"tools","label":"Tools & agent workbenches","shortLabel":"Tools","description":"Runtimes, coding agents, orchestration frameworks, and operator surfaces where AI work actually happens."},{"id":"context","label":"Context engineering & memory","shortLabel":"Context","description":"The supply chain that turns documents, code, web pages, tools, and knowledge graphs into usable agent context."},{"id":"ui","label":"LLM-UI & generated interfaces","shortLabel":"LLM-UI","description":"Interfaces assembled around model output, task state, validation, and human correction."},{"id":"reliability","label":"Evals, specs & reliability","shortLabel":"Reliability","description":"The control plane: specifications, tests, protocols, structured output, security boundaries, and sandboxes."},{"id":"infrastructure","label":"Model & serving infrastructure","shortLabel":"Infrastructure","description":"Serving throughput, local training, cache economics, and the compute layer underneath agent systems."}],"projects":[{"rank":1,"repo":"openclaw/openclaw","url":"https://github.com/openclaw/openclaw","stars":383871,"forks":80650,"theme":"tools","area":"Personal agent runtime","exportHits":2,"why":"Its rapid adoption shows demand for a persistent, cross-platform agent runtime that connects models, tools, channels, and personal workflows."},{"rank":2,"repo":"obra/superpowers","url":"https://github.com/obra/superpowers","stars":259693,"forks":23156,"theme":"tools","area":"Agent skills methodology","exportHits":1,"why":"Reusable skills are becoming an engineering methodology: agents load bounded procedures instead of improvising every workflow from a prompt."},{"rank":3,"repo":"NousResearch/hermes-agent","url":"https://github.com/NousResearch/hermes-agent","stars":219173,"forks":41543,"theme":"tools","area":"Learning agent runtime","exportHits":8,"why":"Hermes makes learning from completed work part of the runtime, turning successful procedures into durable capabilities."},{"rank":4,"repo":"multica-ai/andrej-karpathy-skills","url":"https://github.com/multica-ai/andrej-karpathy-skills","stars":195659,"forks":20145,"theme":"reliability","area":"Coding behavior contract","exportHits":2,"why":"A compact CLAUDE.md derived from recurring coding-agent failures shows how operational lessons are becoming reusable behavioral contracts."},{"rank":5,"repo":"mattpocock/skills","url":"https://github.com/mattpocock/skills","stars":184268,"forks":15771,"theme":"tools","area":"Engineering skills","exportHits":1,"why":"Production-tested engineering procedures packaged as skills make expert workflows portable across agent clients and teams."},{"rank":6,"repo":"microsoft/markitdown","url":"https://github.com/microsoft/markitdown","stars":168372,"forks":12145,"theme":"context","area":"Document ingestion","exportHits":1,"why":"File-to-Markdown conversion is becoming the dependable ingestion layer agents need before retrieval, synthesis, and audit trails."},{"rank":7,"repo":"firecrawl/firecrawl","url":"https://github.com/firecrawl/firecrawl","stars":154693,"forks":8811,"theme":"context","area":"Web extraction","exportHits":1,"why":"Web extraction is moving from scraping glue to an agent-ready primitive for monitoring, RAG, and newsroom intake."},{"rank":8,"repo":"langgenius/dify","url":"https://github.com/langgenius/dify","stars":149881,"forks":23619,"theme":"tools","area":"Agent app platform","exportHits":1,"why":"Dify packages workflows, RAG, tools, and deployment into a shared product surface instead of leaving teams with isolated prompts."},{"rank":9,"repo":"langchain-ai/langchain","url":"https://github.com/langchain-ai/langchain","stars":142379,"forks":23693,"theme":"tools","area":"Agent framework","exportHits":1,"why":"LangChain remains a central abstraction layer for models, tools, retrieval, and production agent plumbing."},{"rank":10,"repo":"x1xhlol/system-prompts-and-models-of-ai-tools","url":"https://github.com/x1xhlol/system-prompts-and-models-of-ai-tools","stars":142204,"forks":34820,"theme":"reliability","area":"Harness transparency","exportHits":5,"why":"The collection exposes how production AI tools combine prompts, tool contracts, and hidden orchestration, making harness design inspectable."},{"rank":11,"repo":"anthropics/claude-code","url":"https://github.com/anthropics/claude-code","stars":138773,"forks":22266,"theme":"tools","area":"Coding agent","exportHits":1,"why":"Terminal-native coding agents turn codebase work into a tool-executing, reviewable loop instead of a chat transcript."},{"rank":12,"repo":"Shubhamsaboo/awesome-llm-apps","url":"https://github.com/Shubhamsaboo/awesome-llm-apps","stars":126567,"forks":18669,"theme":"tools","area":"Reference apps","exportHits":4,"why":"A large runnable example index shows which RAG, agent, and workflow patterns developers actually clone and remix."},{"rank":13,"repo":"garrytan/gstack","url":"https://github.com/garrytan/gstack","stars":123801,"forks":18544,"theme":"tools","area":"Role-based agent stack","exportHits":2,"why":"Opinionated specialist roles package product, design, engineering, release, and QA work into a repeatable agent operating model."},{"rank":14,"repo":"github/spec-kit","url":"https://github.com/github/spec-kit","stars":123353,"forks":10999,"theme":"reliability","area":"Specification workflow","exportHits":1,"why":"Spec-driven development makes requirements, plans, and acceptance checks the control surface before agents write code."},{"rank":15,"repo":"browser-use/browser-use","url":"https://github.com/browser-use/browser-use","stars":106204,"forks":11674,"theme":"tools","area":"Browser agents","exportHits":1,"why":"Browser automation is becoming a first-class capability layer for agents that inspect and act on live web apps."},{"rank":16,"repo":"google-gemini/gemini-cli","url":"https://github.com/google-gemini/gemini-cli","stars":106132,"forks":14299,"theme":"tools","area":"Coding agent","exportHits":1,"why":"A model-backed CLI with tool access and MCP support fits the shift from model choice to routable development harnesses."},{"rank":17,"repo":"openai/codex","url":"https://github.com/openai/codex","stars":100798,"forks":15101,"theme":"tools","area":"Coding agent","exportHits":1,"why":"Codex makes local repository work goal-scoped, command-aware, and evidence-oriented."},{"rank":18,"repo":"karpathy/autoresearch","url":"https://github.com/karpathy/autoresearch","stars":91899,"forks":13142,"theme":"reliability","area":"Autonomous research loop","exportHits":1,"why":"A bounded single-GPU research loop turns hypothesis, experiment, measurement, and iteration into an inspectable autonomous workflow."},{"rank":19,"repo":"punkpeye/awesome-mcp-servers","url":"https://github.com/punkpeye/awesome-mcp-servers","stars":91215,"forks":13439,"theme":"context","area":"Tool ecosystem","exportHits":1,"why":"MCP server catalogs show how quickly agent capabilities are becoming installable infrastructure."},{"rank":20,"repo":"modelcontextprotocol/servers","url":"https://github.com/modelcontextprotocol/servers","stars":88797,"forks":11277,"theme":"context","area":"Tool protocol","exportHits":1,"why":"The reference server collection makes tools and context sources portable across agent clients."},{"rank":21,"repo":"DietrichGebert/ponytail","url":"https://github.com/DietrichGebert/ponytail","stars":88335,"forks":4835,"theme":"reliability","area":"Agent code restraint","exportHits":1,"why":"Encoding senior-engineer restraint into agent behavior addresses the growing cost of unnecessary abstractions and generated-code sprawl."},{"rank":22,"repo":"vllm-project/vllm","url":"https://github.com/vllm-project/vllm","stars":86931,"forks":19742,"theme":"infrastructure","area":"Model serving","exportHits":1,"why":"High-throughput serving keeps model routing and self-hosted inference economically plausible."},{"rank":23,"repo":"bytedance/deer-flow","url":"https://github.com/bytedance/deer-flow","stars":77658,"forks":10573,"theme":"tools","area":"Long-horizon agent","exportHits":3,"why":"Sandboxes, memory, tools, skills, and subagents are converging into one harness for tasks that run for minutes or hours."},{"rank":24,"repo":"openai/openai-cookbook","url":"https://github.com/openai/openai-cookbook","stars":74820,"forks":12656,"theme":"tools","area":"Recipes & harnesses","exportHits":1,"why":"Official examples turn API features into reusable patterns for evaluation, tool use, retrieval, and multimodal work."},{"rank":25,"repo":"unslothai/unsloth","url":"https://github.com/unslothai/unsloth","stars":68764,"forks":6182,"theme":"infrastructure","area":"Local models","exportHits":1,"why":"Faster local training and inference give teams a cheaper private lane for model experiments."},{"rank":26,"repo":"cline/cline","url":"https://github.com/cline/cline","stars":64955,"forks":6975,"theme":"tools","area":"Coding agent SDK","exportHits":1,"why":"Cline's move across IDE, CLI, and SDK surfaces shows coding agents becoming embeddable infrastructure rather than one editor feature."},{"rank":27,"repo":"docling-project/docling","url":"https://github.com/docling-project/docling","stars":63653,"forks":4508,"theme":"context","area":"Document ingestion","exportHits":1,"why":"Production context pipelines need structure-preserving conversion for PDFs and office documents before retrieval can be trustworthy."},{"rank":28,"repo":"warpdotdev/warp","url":"https://github.com/warpdotdev/warp","stars":63580,"forks":5319,"theme":"tools","area":"Agentic terminal","exportHits":1,"why":"The terminal is becoming an operator cockpit where agents, commands, project context, and feedback loops meet."},{"rank":29,"repo":"colbymchenry/codegraph","url":"https://github.com/colbymchenry/codegraph","stars":61941,"forks":3882,"theme":"context","area":"Local code graph","exportHits":1,"why":"A continuously synchronized local code graph reduces repeated repository scans and gives coding agents durable structural context."},{"rank":30,"repo":"Panniantong/Agent-Reach","url":"https://github.com/Panniantong/Agent-Reach","stars":60104,"forks":4825,"theme":"context","area":"Web and social context","exportHits":2,"why":"One CLI for difficult web and social sources makes external evidence retrieval a reusable agent capability rather than per-source scraping glue."},{"rank":31,"repo":"microsoft/autogen","url":"https://github.com/microsoft/autogen","stars":59910,"forks":9019,"theme":"tools","area":"Multi-agent orchestration","exportHits":1,"why":"AutoGen helped normalize agentic programs as explicit conversations, tools, and coordination patterns."},{"rank":32,"repo":"crewAIInc/crewAI","url":"https://github.com/crewAIInc/crewAI","stars":55994,"forks":7925,"theme":"tools","area":"Multi-agent orchestration","exportHits":1,"why":"Crew-style roles remain a practical entry point for decomposing agent work into specialist responsibilities."},{"rank":33,"repo":"mvanhorn/last30days-skill","url":"https://github.com/mvanhorn/last30days-skill","stars":53237,"forks":4607,"theme":"context","area":"Recent-source research","exportHits":1,"why":"A bounded skill for recent cross-platform research packages source discovery, recency, and synthesis into a repeatable evidence workflow."},{"rank":34,"repo":"aaif-goose/goose","url":"https://github.com/aaif-goose/goose","stars":51518,"forks":5668,"theme":"tools","area":"Operations agent","exportHits":1,"why":"Goose represents the open-source move from suggestions to installed, executing, tool-using desktop and code agents."},{"rank":35,"repo":"anthropics/claude-cookbooks","url":"https://github.com/anthropics/claude-cookbooks","stars":49399,"forks":5857,"theme":"tools","area":"Applied recipes","exportHits":5,"why":"Cookbooks turn model capabilities into runnable patterns for tool use, retrieval, multimodal work, and production integration."},{"rank":36,"repo":"chenglou/pretext","url":"https://github.com/chenglou/pretext","stars":49288,"forks":2722,"theme":"ui","area":"Programmatic text layout","exportHits":1,"why":"Fast deterministic text measurement gives generative interfaces a layout primitive agents can calculate and render instead of approximating."},{"rank":37,"repo":"pbakaus/impeccable","url":"https://github.com/pbakaus/impeccable","stars":49047,"forks":2860,"theme":"ui","area":"Design language for agents","exportHits":1,"why":"An explicit design language helps coding agents reason about visual tradeoffs instead of reproducing generic component patterns."},{"rank":38,"repo":"abhigyanpatwari/GitNexus","url":"https://github.com/abhigyanpatwari/GitNexus","stars":44565,"forks":4944,"theme":"context","area":"Client-side code intelligence","exportHits":1,"why":"A browser-local repository graph combines privacy, structural navigation, and Graph RAG without requiring a hosted indexing service."},{"rank":39,"repo":"agno-agi/agno","url":"https://github.com/agno-agi/agno","stars":41369,"forks":5682,"theme":"tools","area":"Agent platform","exportHits":1,"why":"Agno points toward full agent platforms with models, memory, tools, teams, and observability in one stack."},{"rank":40,"repo":"vercel-labs/agent-browser","url":"https://github.com/vercel-labs/agent-browser","stars":39063,"forks":2553,"theme":"tools","area":"Browser automation CLI","exportHits":1,"why":"A browser CLI designed around agent calls makes web interaction scriptable, inspectable, and easier to embed in coding and operations loops."},{"rank":41,"repo":"langchain-ai/langgraph","url":"https://github.com/langchain-ai/langgraph","stars":37905,"forks":6366,"theme":"tools","area":"Stateful agents","exportHits":1,"why":"Graph-shaped state, checkpoints, and control flow are becoming the durable architecture behind long-running agents."},{"rank":42,"repo":"google/langextract","url":"https://github.com/google/langextract","stars":37756,"forks":2611,"theme":"context","area":"Grounded extraction","exportHits":1,"why":"Structured extraction with precise source grounding connects probabilistic models to auditable data pipelines."},{"rank":43,"repo":"heygen-com/hyperframes","url":"https://github.com/heygen-com/hyperframes","stars":37010,"forks":3489,"theme":"ui","area":"Agent-native video","exportHits":1,"why":"HTML-to-video gives agents a code-native visual medium they can generate, inspect, revise, and render deterministically."},{"rank":44,"repo":"github/awesome-copilot","url":"https://github.com/github/awesome-copilot","stars":36951,"forks":4624,"theme":"tools","area":"Agent configuration","exportHits":1,"why":"Shared instructions, skills, and agent configurations show development practice moving from private prompting into reusable operational assets."},{"rank":45,"repo":"tinyhumansai/openhuman","url":"https://github.com/tinyhumansai/openhuman","stars":35271,"forks":3461,"theme":"tools","area":"Personal agent harness","exportHits":1,"why":"Local-first personal memory plus fleet orchestration points toward assistants becoming durable operating environments rather than isolated chats."},{"rank":46,"repo":"anthropics/financial-services","url":"https://github.com/anthropics/financial-services","stars":33731,"forks":4985,"theme":"tools","area":"Domain agent modules","exportHits":1,"why":"Reference agents, connectors, and skills for financial work show how professional teams can be packaged as inspectable domain capability modules."},{"rank":47,"repo":"anthropics/claude-plugins-official","url":"https://github.com/anthropics/claude-plugins-official","stars":32526,"forks":3644,"theme":"tools","area":"Plugin ecosystem","exportHits":2,"why":"Curated plugins turn domain workflows and integrations into installable capability packages with explicit ownership."},{"rank":48,"repo":"Yeachan-Heo/oh-my-codex","url":"https://github.com/Yeachan-Heo/oh-my-codex","stars":32195,"forks":2496,"theme":"tools","area":"Codex agent teams","exportHits":1,"why":"Hooks, agent teams, and operator HUDs extend a coding agent into a configurable multi-agent workbench."},{"rank":49,"repo":"lightpanda-io/browser","url":"https://github.com/lightpanda-io/browser","stars":32094,"forks":1431,"theme":"infrastructure","area":"Agent browser engine","exportHits":1,"why":"A headless browser built for automation treats browser execution cost and throughput as first-class agent infrastructure."},{"rank":50,"repo":"topoteretes/cognee","url":"https://github.com/topoteretes/cognee","stars":29190,"forks":2772,"theme":"context","area":"Agent memory","exportHits":1,"why":"Cognee turns memory into a self-hosted knowledge graph layer, matching New Runtime's shared-context thesis."},{"rank":51,"repo":"getzep/graphiti","url":"https://github.com/getzep/graphiti","stars":29093,"forks":2933,"theme":"context","area":"Temporal memory","exportHits":1,"why":"Temporal knowledge graphs make memory updateable, queryable, and less tied to one chat window."},{"rank":52,"repo":"supermemoryai/supermemory","url":"https://github.com/supermemoryai/supermemory","stars":28563,"forks":2482,"theme":"context","area":"Memory and context engine","exportHits":1,"why":"A local-capable memory API separates persistent context from any one model, client, or conversation."},{"rank":53,"repo":"openai/openai-agents-python","url":"https://github.com/openai/openai-agents-python","stars":28107,"forks":4365,"theme":"tools","area":"Agent SDK","exportHits":2,"why":"The SDK packages tools, handoffs, tracing, and guardrails as code instead of one-off prompt structure."},{"rank":54,"repo":"opendataloader-project/opendataloader-pdf","url":"https://github.com/opendataloader-project/opendataloader-pdf","stars":27724,"forks":2659,"theme":"context","area":"PDF ingestion","exportHits":1,"why":"AI-ready PDF parsing is becoming a dedicated infrastructure layer because document layout and provenance cannot be recovered after naive text extraction."},{"rank":55,"repo":"yamadashy/repomix","url":"https://github.com/yamadashy/repomix","stars":27332,"forks":1446,"theme":"context","area":"Code context","exportHits":1,"why":"Packing repositories into AI-readable context remains a practical bridge between large codebases and coding agents."},{"rank":56,"repo":"vercel-labs/skills","url":"https://github.com/vercel-labs/skills","stars":27049,"forks":2279,"theme":"tools","area":"Skills package manager","exportHits":1,"why":"A cross-client installer and discovery layer signals that agent skills are becoming a package ecosystem rather than copied folders."},{"rank":57,"repo":"PrefectHQ/fastmcp","url":"https://github.com/PrefectHQ/fastmcp","stars":26783,"forks":2179,"theme":"context","area":"Tool protocol","exportHits":1,"why":"FastMCP lowers the cost of turning internal APIs and scripts into agent-callable tools."},{"rank":58,"repo":"langchain-ai/deepagents","url":"https://github.com/langchain-ai/deepagents","stars":26696,"forks":3736,"theme":"tools","area":"Agent harness","exportHits":2,"why":"DeepAgents packages planning, subagents, file state, and long-running behavior into a batteries-included harness."},{"rank":59,"repo":"Fosowl/agenticSeek","url":"https://github.com/Fosowl/agenticSeek","stars":26673,"forks":2989,"theme":"tools","area":"Local autonomous agent","exportHits":1,"why":"A fully local browsing and coding agent reflects demand for private autonomy without per-task hosted API dependence."},{"rank":60,"repo":"mastra-ai/mastra","url":"https://github.com/mastra-ai/mastra","stars":26468,"forks":2498,"theme":"tools","area":"TypeScript agents","exportHits":1,"why":"Mastra shows the JavaScript stack absorbing agents as production application components."},{"rank":61,"repo":"google-labs-code/design.md","url":"https://github.com/google-labs-code/design.md","stars":26264,"forks":2096,"theme":"ui","area":"Design context standard","exportHits":1,"why":"A portable design-system file gives coding agents durable visual context instead of forcing them to infer taste from screenshots on every task."},{"rank":62,"repo":"openai/symphony","url":"https://github.com/openai/symphony","stars":26168,"forks":2643,"theme":"tools","area":"Autonomous work manager","exportHits":2,"why":"Isolated implementation runs move the operator from supervising every command to managing goals, evidence, and completed work."},{"rank":63,"repo":"deepset-ai/haystack","url":"https://github.com/deepset-ai/haystack","stars":25987,"forks":2944,"theme":"context","area":"RAG pipelines","exportHits":1,"why":"Haystack keeps context engineering explicit through pipelines, retrieval, routing, memory, and observability."},{"rank":64,"repo":"rohitg00/agentmemory","url":"https://github.com/rohitg00/agentmemory","stars":25662,"forks":2133,"theme":"context","area":"Coding-agent memory","exportHits":1,"why":"Shared persistent memory across coding agents reduces repeated onboarding and makes context continuity measurable against real tasks."},{"rank":65,"repo":"a2aproject/A2A","url":"https://github.com/a2aproject/A2A","stars":24962,"forks":2524,"theme":"reliability","area":"Agent protocol","exportHits":1,"why":"Agent-to-agent protocols matter as soon as agents become opaque services owned by different products or teams."},{"rank":66,"repo":"openai/skills","url":"https://github.com/openai/skills","stars":24108,"forks":1638,"theme":"tools","area":"Codex skills catalog","exportHits":1,"why":"An official skills catalog moves reusable Codex procedures into a discoverable, versionable capability layer."},{"rank":67,"repo":"agentskills/agentskills","url":"https://github.com/agentskills/agentskills","stars":23368,"forks":1561,"theme":"reliability","area":"Agent skills standard","exportHits":1,"why":"A shared skill specification makes procedures portable across agent clients and gives teams a stable contract for capability packaging."},{"rank":68,"repo":"ogulcancelik/herdr","url":"https://github.com/ogulcancelik/herdr","stars":19886,"forks":1303,"theme":"tools","area":"Agent terminal multiplexer","exportHits":1,"why":"A terminal multiplexer with agent-aware states makes parallel coding sessions observable as work rather than anonymous panes."},{"rank":69,"repo":"mksglu/context-mode","url":"https://github.com/mksglu/context-mode","stars":19225,"forks":1352,"theme":"context","area":"Context-window optimization","exportHits":1,"why":"Sandboxing tool output and routing context through MCP and hooks turns token pressure into an explicit systems optimization problem."},{"rank":70,"repo":"allenai/olmocr","url":"https://github.com/allenai/olmocr","stars":19161,"forks":1578,"theme":"context","area":"Document linearization","exportHits":1,"why":"High-quality PDF linearization improves the training and retrieval substrate before an LLM ever sees the document."},{"rank":71,"repo":"openai/evals","url":"https://github.com/openai/evals","stars":18970,"forks":3031,"theme":"reliability","area":"Evaluation","exportHits":1,"why":"Evals are moving from model benchmarks into product infrastructure for judging complete AI systems."},{"rank":72,"repo":"pydantic/pydantic-ai","url":"https://github.com/pydantic/pydantic-ai","stars":18767,"forks":2403,"theme":"tools","area":"Typed agent framework","exportHits":1,"why":"Typed dependencies, structured outputs, tools, and eval integrations bring agent construction closer to ordinary production Python engineering."},{"rank":73,"repo":"baidu/Unlimited-OCR","url":"https://github.com/baidu/Unlimited-OCR","stars":18086,"forks":1706,"theme":"context","area":"Long-document OCR","exportHits":1,"why":"One-shot long-horizon document parsing challenges page-by-page pipelines and reduces the reconstruction burden in document context systems."},{"rank":74,"repo":"camel-ai/camel","url":"https://github.com/camel-ai/camel","stars":17465,"forks":2016,"theme":"tools","area":"Multi-agent research","exportHits":2,"why":"CAMEL keeps the research thread alive around scaling laws, roles, societies, and cooperation among agents."},{"rank":75,"repo":"microsoft/agent-lightning","url":"https://github.com/microsoft/agent-lightning","stars":17412,"forks":1525,"theme":"reliability","area":"Agent optimization","exportHits":1,"why":"Agent training is shifting from prompt tweaking toward trace-level optimization of complete trajectories and tool decisions."},{"rank":76,"repo":"confident-ai/deepeval","url":"https://github.com/confident-ai/deepeval","stars":17065,"forks":1697,"theme":"reliability","area":"Evaluation","exportHits":1,"why":"DeepEval makes LLM and agent quality measurable in CI-style loops instead of relying on manual prompt taste."},{"rank":77,"repo":"rowboatlabs/rowboat","url":"https://github.com/rowboatlabs/rowboat","stars":16734,"forks":1664,"theme":"tools","area":"Memory-backed coworker","exportHits":1,"why":"An open coworker that combines tools with persistent memory reflects the shift from chat assistants to accumulating work relationships."},{"rank":78,"repo":"vercel-labs/json-render","url":"https://github.com/vercel-labs/json-render","stars":15747,"forks":849,"theme":"ui","area":"Generative UI","exportHits":1,"why":"Structured UI output lets model answers become inspectable interfaces instead of unstructured prose blobs."},{"rank":79,"repo":"MemoriLabs/Memori","url":"https://github.com/MemoriLabs/Memori","stars":15644,"forks":2874,"theme":"context","area":"Production memory","exportHits":1,"why":"Agent-native memory that writes structured state into existing data infrastructure matches the move from chat history to governed organizational memory."},{"rank":80,"repo":"dottxt-ai/outlines","url":"https://github.com/dottxt-ai/outlines","stars":15201,"forks":807,"theme":"reliability","area":"Structured generation","exportHits":1,"why":"Grammar and schema-constrained outputs reduce the gap between probabilistic generation and software contracts."},{"rank":81,"repo":"google/skills","url":"https://github.com/google/skills","stars":15176,"forks":1176,"theme":"tools","area":"Provider skill library","exportHits":1,"why":"Provider-maintained procedures for products and technologies make operational knowledge directly loadable by agents."},{"rank":82,"repo":"eigent-ai/eigent","url":"https://github.com/eigent-ai/eigent","stars":14630,"forks":1734,"theme":"tools","area":"Agent desktop","exportHits":1,"why":"Open agent desktops are becoming operator workspaces where local tools, files, and multi-step execution meet."},{"rank":83,"repo":"NevaMind-AI/memU","url":"https://github.com/NevaMind-AI/memU","stars":14053,"forks":1043,"theme":"context","area":"Cross-agent memory","exportHits":1,"why":"Memory shared across agents addresses the fragmentation caused when every client maintains a separate understanding of the user."},{"rank":84,"repo":"NVIDIA/SkillSpector","url":"https://github.com/NVIDIA/SkillSpector","stars":13567,"forks":1123,"theme":"reliability","area":"Skill security","exportHits":1,"why":"As skills become executable supply-chain artifacts, static inspection for malicious instructions and risky behavior becomes mandatory."},{"rank":85,"repo":"langchain-ai/openwiki","url":"https://github.com/langchain-ai/openwiki","stars":12975,"forks":896,"theme":"context","area":"Codebase documentation","exportHits":2,"why":"Agent-maintained repository documentation creates a durable context layer that stays closer to the changing code."},{"rank":86,"repo":"StarTrail-org/LEANN","url":"https://github.com/StarTrail-org/LEANN","stars":12719,"forks":1148,"theme":"context","area":"Local private RAG","exportHits":1,"why":"Storage-efficient local retrieval makes private personal and edge-device RAG practical without a heavyweight vector database."},{"rank":87,"repo":"InsForge/InsForge","url":"https://github.com/InsForge/InsForge","stars":12433,"forks":1086,"theme":"tools","area":"Backend for coding agents","exportHits":1,"why":"A backend surface designed for agentic coding lets agents provision data, auth, storage, compute, and deployment through one explicit contract."},{"rank":88,"repo":"zilliztech/claude-context","url":"https://github.com/zilliztech/claude-context","stars":12188,"forks":904,"theme":"context","area":"Code search MCP","exportHits":1,"why":"Repository-wide semantic search exposed through MCP gives multiple coding clients a shared context service."},{"rank":89,"repo":"tambo-ai/tambo","url":"https://github.com/tambo-ai/tambo","stars":11163,"forks":559,"theme":"ui","area":"Generative UI","exportHits":1,"why":"A React generative-UI SDK points to applications where the interface is assembled around the task state."},{"rank":90,"repo":"cocoindex-io/cocoindex","url":"https://github.com/cocoindex-io/cocoindex","stars":11008,"forks":848,"theme":"context","area":"Incremental context","exportHits":1,"why":"Incremental indexing keeps long-running agent context fresh without paying for full reprocessing after every source change."},{"rank":91,"repo":"LMCache/LMCache","url":"https://github.com/LMCache/LMCache","stars":10824,"forks":1603,"theme":"infrastructure","area":"KV-cache","exportHits":1,"why":"KV-cache infrastructure turns long context and repeated workloads into a serving optimization problem."},{"rank":92,"repo":"MoonshotAI/kimi-cli","url":"https://github.com/MoonshotAI/kimi-cli","stars":10656,"forks":1240,"theme":"tools","area":"Coding agent CLI","exportHits":1,"why":"Another capable terminal agent reinforces the trend toward model-swappable development harnesses with tools and repository access."},{"rank":93,"repo":"TencentCloud/CubeSandbox","url":"https://github.com/TencentCloud/CubeSandbox","stars":10609,"forks":946,"theme":"reliability","area":"Agent sandbox","exportHits":1,"why":"Isolated execution environments are required before agents can safely run code, browsers, or external tools."},{"rank":94,"repo":"0x4m4/hexstrike-ai","url":"https://github.com/0x4m4/hexstrike-ai","stars":10443,"forks":2175,"theme":"reliability","area":"Agent security","exportHits":1,"why":"Offensive MCP tooling signals that agent tool access is both a capability frontier and a security boundary."},{"rank":95,"repo":"github/copilot-sdk","url":"https://github.com/github/copilot-sdk","stars":10042,"forks":1355,"theme":"tools","area":"Embedded agent SDK","exportHits":1,"why":"An SDK turns the coding agent from a destination product into a capability that other applications can embed."},{"rank":96,"repo":"microsoft/magentic-ui","url":"https://github.com/microsoft/magentic-ui","stars":9976,"forks":1003,"theme":"ui","area":"Agent interface","exportHits":1,"why":"Browser and local-file agents need an operator interface, not just a backend chain of tool calls."},{"rank":97,"repo":"shadcn/improve","url":"https://github.com/shadcn/improve","stars":8598,"forks":372,"theme":"reliability","area":"Model-tiered code audit","exportHits":2,"why":"Separating expensive architectural review from cheaper execution models turns model routing into a concrete engineering workflow."},{"rank":98,"repo":"idosal/git-mcp","url":"https://github.com/idosal/git-mcp","stars":8278,"forks":735,"theme":"context","area":"Repository context","exportHits":1,"why":"A remote MCP context layer gives agents source-grounded access to repositories without manually packing every codebase into the prompt."},{"rank":99,"repo":"pydantic/monty","url":"https://github.com/pydantic/monty","stars":7933,"forks":394,"theme":"reliability","area":"Secure agent interpreter","exportHits":1,"why":"A minimal Rust-based Python interpreter gives agents a deliberately bounded execution surface instead of a full host runtime."},{"rank":100,"repo":"google-labs-code/stitch-skills","url":"https://github.com/google-labs-code/stitch-skills","stars":7774,"forks":916,"theme":"ui","area":"Design agent skills","exportHits":1,"why":"Design procedures packaged as portable skills connect generative UI tools to repeatable production workflows."},{"rank":101,"repo":"evidentlyai/evidently","url":"https://github.com/evidentlyai/evidently","stars":7742,"forks":886,"theme":"reliability","area":"LLM observability","exportHits":1,"why":"Evaluation, testing, and monitoring are converging into one operational loop for AI systems after deployment."},{"rank":102,"repo":"teamchong/pxpipe","url":"https://github.com/teamchong/pxpipe","stars":6637,"forks":565,"theme":"context","area":"Visual context compression","exportHits":2,"why":"Rendering bulky text context as images explores a new tradeoff between multimodal input capacity, token cost, and recoverability."},{"rank":103,"repo":"GoogleCloudPlatform/agent-starter-pack","url":"https://github.com/GoogleCloudPlatform/agent-starter-pack","stars":6520,"forks":1497,"theme":"tools","area":"Production templates","exportHits":1,"why":"Agent templates with CI/CD, evaluation, and observability compress the path from demo to governed cloud deployment."},{"rank":104,"repo":"airweave-ai/airweave","url":"https://github.com/airweave-ai/airweave","stars":6494,"forks":816,"theme":"context","area":"Context retrieval","exportHits":1,"why":"A dedicated retrieval layer separates continuously changing source synchronization from the agent application itself."},{"rank":105,"repo":"steipete/agent-scripts","url":"https://github.com/steipete/agent-scripts","stars":6451,"forks":535,"theme":"tools","area":"Operational scripts","exportHits":5,"why":"Small reusable scripts make recurring maintenance and review workflows callable by agents instead of rediscovered ad hoc."},{"rank":106,"repo":"max-sixty/worktrunk","url":"https://github.com/max-sixty/worktrunk","stars":5965,"forks":208,"theme":"tools","area":"Parallel-agent worktrees","exportHits":1,"why":"Agent-oriented worktree management makes isolated parallel coding sessions reproducible and easier to supervise."},{"rank":107,"repo":"1weiho/open-slide","url":"https://github.com/1weiho/open-slide","stars":5953,"forks":416,"theme":"ui","area":"Agent-native slides","exportHits":1,"why":"A code-native slide framework gives agents structured layout, components, and renderable output rather than opaque presentation files."},{"rank":108,"repo":"MinishLab/semble","url":"https://github.com/MinishLab/semble","stars":5684,"forks":237,"theme":"context","area":"Code search","exportHits":1,"why":"Token-efficient structural code search reduces the context tax of finding the right repository evidence before editing."},{"rank":109,"repo":"agent-infra/sandbox","url":"https://github.com/agent-infra/sandbox","stars":5473,"forks":491,"theme":"reliability","area":"Integrated agent sandbox","exportHits":1,"why":"Combining browser, shell, files, MCP, and an editor in one container gives agent runs a bounded and reproducible execution environment."},{"rank":110,"repo":"google/agents-cli","url":"https://github.com/google/agents-cli","stars":5326,"forks":563,"theme":"tools","area":"Agent delivery CLI","exportHits":1,"why":"Creating, evaluating, and deploying agents through one CLI makes the lifecycle a reproducible engineering workflow."},{"rank":111,"repo":"vercel-labs/zerolang","url":"https://github.com/vercel-labs/zerolang","stars":5224,"forks":341,"theme":"reliability","area":"Agent programming language","exportHits":1,"why":"A language built around structured diagnostics gives coding agents machine-readable repair signals instead of prose compiler output."},{"rank":112,"repo":"jgraph/drawio-mcp","url":"https://github.com/jgraph/drawio-mcp","stars":4947,"forks":310,"theme":"ui","area":"Streaming diagram UI","exportHits":2,"why":"Streaming diagram primitives through MCP turns model-generated structure into an inspectable visual work surface while it is being built."},{"rank":113,"repo":"mistralai/mistral-vibe","url":"https://github.com/mistralai/mistral-vibe","stars":4724,"forks":608,"theme":"tools","area":"Coding agent CLI","exportHits":1,"why":"A minimal coding CLI from a model provider shows the harness layer becoming as strategic as the model endpoint."},{"rank":114,"repo":"mixedbread-ai/mgrep","url":"https://github.com/mixedbread-ai/mgrep","stars":4306,"forks":171,"theme":"context","area":"Multimodal search","exportHits":1,"why":"Semantic grep across code, images, and documents offers agents one retrieval primitive across heterogeneous project context."},{"rank":115,"repo":"rivet-dev/agentos","url":"https://github.com/rivet-dev/agentos","stars":4039,"forks":200,"theme":"tools","area":"Embedded agent runtime","exportHits":2,"why":"A library-level runtime using isolates lets applications host bounded agents inside existing backends without a separate SaaS control plane."},{"rank":116,"repo":"cursor/cookbook","url":"https://github.com/cursor/cookbook","stars":3974,"forks":470,"theme":"tools","area":"Coding-agent recipes","exportHits":1,"why":"Official workflow recipes expose the operating patterns around coding agents, not only the editor feature set."},{"rank":117,"repo":"google-gemini/gemini-skills","url":"https://github.com/google-gemini/gemini-skills","stars":3840,"forks":388,"theme":"tools","area":"Model skills","exportHits":2,"why":"Provider-maintained skills make model and SDK expertise loadable as procedures instead of copied documentation."},{"rank":118,"repo":"BuilderIO/skills","url":"https://github.com/BuilderIO/skills","stars":3754,"forks":190,"theme":"tools","area":"Visual engineering skills","exportHits":1,"why":"Portable planning and visual-recap skills make graphical artifacts part of the coding-agent coordination loop."},{"rank":119,"repo":"memodb-io/Acontext","url":"https://github.com/memodb-io/Acontext","stars":3584,"forks":326,"theme":"context","area":"Skill memory","exportHits":1,"why":"Treating skills as memory connects reusable procedures with the state and evidence agents need to apply them correctly."},{"rank":120,"repo":"openai/openai-agents-js","url":"https://github.com/openai/openai-agents-js","stars":3448,"forks":868,"theme":"tools","area":"JavaScript agent SDK","exportHits":1,"why":"A JavaScript SDK brings tools, handoffs, tracing, and voice agents into the dominant web application runtime."},{"rank":121,"repo":"oceanbase/seekdb","url":"https://github.com/oceanbase/seekdb","stars":2837,"forks":311,"theme":"context","area":"Agent storage","exportHits":1,"why":"Unifying vector, text, structured, and semi-structured retrieval reduces the storage fragmentation behind agent memory."},{"rank":122,"repo":"simonw/files-to-prompt","url":"https://github.com/simonw/files-to-prompt","stars":2764,"forks":176,"theme":"context","area":"File context packing","exportHits":1,"why":"A small deterministic context packer remains useful because inspectable preprocessing is often preferable to opaque ingestion."},{"rank":123,"repo":"Agent-Field/agentfield","url":"https://github.com/Agent-Field/agentfield","stars":2431,"forks":385,"theme":"tools","area":"Agent control plane","exportHits":2,"why":"Treating agents as routable APIs and microservices puts memory, asynchronous execution, scaling, and audit into one control plane."},{"rank":124,"repo":"cursor/plugins","url":"https://github.com/cursor/plugins","stars":2427,"forks":187,"theme":"tools","area":"Coding-agent plugins","exportHits":1,"why":"An official plugin specification makes coding-agent extensions installable and governed instead of hidden in personal configuration."},{"rank":125,"repo":"openai/openai-apps-sdk-examples","url":"https://github.com/openai/openai-apps-sdk-examples","stars":2282,"forks":516,"theme":"ui","area":"Embedded agent UI","exportHits":1,"why":"Runnable Apps SDK examples show how model responses become interactive, stateful product surfaces instead of plain text."},{"rank":126,"repo":"benchflow-ai/skillsbench","url":"https://github.com/benchflow-ai/skillsbench","stars":1566,"forks":343,"theme":"reliability","area":"Skill evaluation","exportHits":1,"why":"Skills need behavioral benchmarks that measure both the procedure and the agent's ability to invoke it correctly."},{"rank":127,"repo":"chunkhound/chunkhound","url":"https://github.com/chunkhound/chunkhound","stars":1367,"forks":114,"theme":"context","area":"Local code intelligence","exportHits":1,"why":"Local-first codebase intelligence gives agents structural retrieval without exporting private repositories to a hosted index."},{"rank":128,"repo":"coinbase/agentkit","url":"https://github.com/coinbase/agentkit","stars":1275,"forks":754,"theme":"tools","area":"Agent wallet toolkit","exportHits":1,"why":"Wallet primitives let agents transact under explicit tool contracts, pushing identity, permissions, and audit into the agent economy."},{"rank":129,"repo":"langchain-ai/openevals","url":"https://github.com/langchain-ai/openevals","stars":1126,"forks":106,"theme":"reliability","area":"Evaluation library","exportHits":1,"why":"Ready-made evaluators lower the cost of putting repeatable quality checks around LLM and agent behavior."},{"rank":130,"repo":"AI-Builder-Club/skills","url":"https://github.com/AI-Builder-Club/skills","stars":1036,"forks":140,"theme":"reliability","area":"Loop engineering skills","exportHits":2,"why":"Codebase maps and loop-engineering procedures package structural evidence and repeatable maintenance cycles for coding agents."},{"rank":131,"repo":"SafeAI-Lab-X/ClawKeeper","url":"https://github.com/SafeAI-Lab-X/ClawKeeper","stars":1033,"forks":58,"theme":"reliability","area":"Agent runtime security","exportHits":1,"why":"Skills, plugins, and runtime watchers form a layered security boundary around a tool-using personal agent."},{"rank":132,"repo":"microsoft/work-iq","url":"https://github.com/microsoft/work-iq","stars":947,"forks":109,"theme":"context","area":"Enterprise work context","exportHits":1,"why":"An MCP and CLI surface for organizational work context points toward governed enterprise knowledge becoming directly agent-callable."},{"rank":133,"repo":"openai/realtime-voice-component","url":"https://github.com/openai/realtime-voice-component","stars":878,"forks":113,"theme":"ui","area":"Realtime voice UI","exportHits":1,"why":"A reusable voice component makes low-latency multimodal interaction an embeddable interface primitive rather than a bespoke demo."},{"rank":134,"repo":"CopilotKit/generative-ui","url":"https://github.com/CopilotKit/generative-ui","stars":783,"forks":69,"theme":"ui","area":"Generative UI protocols","exportHits":1,"why":"Examples across AG-UI, A2UI, Open JSON UI, and MCP Apps reveal an emerging protocol layer for model-generated interfaces."},{"rank":135,"repo":"langchain-ai/agent-protocol","url":"https://github.com/langchain-ai/agent-protocol","stars":637,"forks":55,"theme":"reliability","area":"Agent service contract","exportHits":1,"why":"A framework-neutral service contract separates agent infrastructure and lifecycle operations from any one orchestration library."},{"rank":136,"repo":"senlindesign/taste-skill","url":"https://github.com/senlindesign/taste-skill","stars":241,"forks":17,"theme":"ui","area":"Design reasoning skill","exportHits":1,"why":"Extracting design decisions and tradeoffs rather than only tokens gives interface agents a more durable representation of visual intent."},{"rank":137,"repo":"zapier/AutomationBench","url":"https://github.com/zapier/AutomationBench","stars":141,"forks":16,"theme":"reliability","area":"Business-workflow benchmark","exportHits":1,"why":"Realistic business automation tasks test whether agents can complete multi-step work across tools, not merely answer benchmark questions."},{"rank":138,"repo":"OpenDataBox/MemoryData","url":"https://github.com/OpenDataBox/MemoryData","stars":120,"forks":19,"theme":"reliability","area":"Agent memory benchmark","exportHits":1,"why":"A unified benchmark for memory-augmented agents makes storage, extraction, retrieval, and use comparable as a system."},{"rank":139,"repo":"cisco-foundation-ai/fully-automated-prompt-optimization","url":"https://github.com/cisco-foundation-ai/fully-automated-prompt-optimization","stars":99,"forks":12,"theme":"reliability","area":"Pipeline self-optimization","exportHits":1,"why":"Optimizing multi-step LLM chains with coding agents and evals moves improvement from manual prompt editing to a measured loop."},{"rank":140,"repo":"AgentToolkit/altk-evolve","url":"https://github.com/AgentToolkit/altk-evolve","stars":95,"forks":11,"theme":"reliability","area":"Iterative agent improvement","exportHits":1,"why":"Iteration-based self-improvement makes agent changes explicit artifacts that can be compared and evaluated between runs."}],"ecosystem":{"type":"modern_ai_ecosystem","verified_at":"2026-07-30","source_note":"Two owner-supplied reference images normalized against public source pages and official product or repository links.","categories":[{"id":"deployment","label":"Deployment & inference","role":"Run models and AI workloads with an explicit serving, cloud, latency, and cost boundary."},{"id":"llm","label":"LLMs","role":"Provide the model intelligence layer; model choice is one dependency, not the whole system."},{"id":"agentic","label":"Agentic orchestration","role":"Coordinate state, planning, tool calls, handoffs, and multi-agent execution."},{"id":"rag","label":"RAG frameworks","role":"Turn retrieval, grounding, ranking, and generation into inspectable pipelines."},{"id":"embedding","label":"Embeddings","role":"Translate content into a semantic space for search, clustering, and retrieval."},{"id":"data-extraction","label":"Data extraction","role":"Convert web pages, documents, and PDFs into structured, attributable context."},{"id":"mcp","label":"MCP","role":"Expose tools and context sources through a shared protocol rather than client-specific glue."},{"id":"security","label":"AI security","role":"Constrain unsafe input, output, data exposure, and tool execution before production writes."},{"id":"observability","label":"Evaluation & observability","role":"Trace behavior, measure quality, compare versions, and catch regressions across the full loop."},{"id":"memory","label":"Memory","role":"Persist state and knowledge beyond one context window while keeping updates auditable."},{"id":"agents","label":"Agent SDKs","role":"Package tools, handoffs, state, and runtime behavior into application-ready agent components."},{"id":"automation","label":"Automation","role":"Connect model decisions to durable workflows, retries, schedules, and human approval gates."},{"id":"vector-database","label":"Vector databases","role":"Store and query semantic representations with metadata, filtering, and operational controls."}],"production_levels":[{"level":0,"label":"Deployment","role":"Where the workload runs and how inference reaches production."},{"level":1,"label":"Evaluation","role":"How quality, traces, retrieval, and regressions are measured."},{"level":2,"label":"LLMs","role":"The reasoning and generation engines selected for each task."},{"level":3,"label":"Framework","role":"The orchestration and retrieval layer that binds models to application logic."},{"level":4,"label":"Vector DB","role":"The semantic retrieval substrate for indexed knowledge."},{"level":5,"label":"Embeddings","role":"The representation layer that controls semantic similarity."},{"level":6,"label":"Data extraction","role":"The ingestion layer that recovers usable structure and provenance."},{"level":7,"label":"Memory","role":"The persistent context layer beneath individual requests."},{"level":8,"label":"Alignment","role":"The safety, policy, monitoring, and control layer around the complete system."}],"tools":[{"id":"groq","name":"Groq","url":"https://console.groq.com/docs/overview","linkKind":"official-docs","categories":["deployment"],"productionLevels":[0]},{"id":"aws-bedrock","name":"AWS Bedrock","url":"https://docs.aws.amazon.com/bedrock/","linkKind":"official-docs","categories":["deployment"],"productionLevels":[0]},{"id":"google-cloud-vertex-ai","name":"Google Cloud Vertex AI","url":"https://cloud.google.com/vertex-ai/generative-ai/docs/overview","linkKind":"official-docs","categories":["deployment"],"productionLevels":[0]},{"id":"together-ai","name":"Together AI","url":"https://docs.together.ai/docs/introduction","linkKind":"official-docs","categories":["deployment"],"productionLevels":[0]},{"id":"openai-gpt","name":"OpenAI GPT","url":"https://platform.openai.com/docs/models","linkKind":"official-docs","categories":["llm"],"productionLevels":[2]},{"id":"anthropic-claude","name":"Anthropic Claude","url":"https://docs.anthropic.com/en/docs/about-claude/models/overview","linkKind":"official-docs","categories":["llm"],"productionLevels":[2]},{"id":"google-gemini","name":"Google Gemini","url":"https://ai.google.dev/gemini-api/docs/models","linkKind":"official-docs","categories":["llm"],"productionLevels":[2]},{"id":"meta-llama","name":"Meta Llama","url":"https://github.com/meta-llama/llama-models","linkKind":"github","categories":["llm"]},{"id":"mistral-ai","name":"Mistral AI","url":"https://docs.mistral.ai/getting-started/models/models_overview/","linkKind":"official-docs","categories":["llm"]},{"id":"cohere-command","name":"Cohere","url":"https://docs.cohere.com/docs/models","linkKind":"official-docs","categories":["llm"]},{"id":"hugging-face-transformers","name":"Hugging Face","url":"https://github.com/huggingface/transformers","linkKind":"github","categories":["deployment","llm"]},{"id":"ollama","name":"Ollama","url":"https://github.com/ollama/ollama","linkKind":"github","categories":["deployment","llm","embedding"],"productionLevels":[5]},{"id":"vllm","name":"vLLM","url":"https://github.com/vllm-project/vllm","linkKind":"github","categories":["deployment","llm"]},{"id":"kimi","name":"Kimi","url":"https://platform.kimi.ai/docs/overview","linkKind":"official-docs","categories":["llm"],"productionLevels":[2]},{"id":"langgraph","name":"LangGraph","url":"https://github.com/langchain-ai/langgraph","linkKind":"github","categories":["agentic"]},{"id":"crewai","name":"CrewAI","url":"https://github.com/crewAIInc/crewAI","linkKind":"github","categories":["agentic"]},{"id":"microsoft-autogen","name":"Microsoft AutoGen","url":"https://github.com/microsoft/autogen","linkKind":"github","categories":["agentic"]},{"id":"microsoft-agent-framework","name":"Microsoft Agent Framework","url":"https://github.com/microsoft/agent-framework","linkKind":"github","categories":["agentic"]},{"id":"llamaindex","name":"LlamaIndex","url":"https://github.com/run-llama/llama_index","linkKind":"github","categories":["agentic","rag"],"productionLevels":[3]},{"id":"aws-strands-agents","name":"AWS Strands Agents","url":"https://github.com/strands-agents/sdk-python","linkKind":"github","categories":["agentic"]},{"id":"camel","name":"CAMEL","url":"https://github.com/camel-ai/camel","linkKind":"github","categories":["agentic"]},{"id":"agno","name":"Agno","url":"https://github.com/agno-agi/agno","linkKind":"github","categories":["agentic"]},{"id":"langchain","name":"LangChain","url":"https://github.com/langchain-ai/langchain","linkKind":"github","categories":["rag","agents"],"productionLevels":[3]},{"id":"haystack","name":"Haystack","url":"https://github.com/deepset-ai/haystack","linkKind":"github","categories":["rag"],"productionLevels":[3]},{"id":"dspy","name":"DSPy","url":"https://github.com/stanfordnlp/dspy","linkKind":"github","categories":["rag"],"productionLevels":[3]},{"id":"ragflow","name":"RAGFlow","url":"https://github.com/infiniflow/ragflow","linkKind":"github","categories":["rag"]},{"id":"graphrag","name":"GraphRAG","url":"https://github.com/microsoft/graphrag","linkKind":"github","categories":["rag"]},{"id":"unstructured","name":"Unstructured","url":"https://github.com/Unstructured-IO/unstructured","linkKind":"github","categories":["rag","data-extraction"]},{"id":"embedchain","name":"Embedchain","url":"https://github.com/mem0ai/embedchain","linkKind":"github","categories":["rag"]},{"id":"openai-embeddings","name":"OpenAI Embeddings","url":"https://platform.openai.com/docs/guides/embeddings","linkKind":"official-docs","categories":["embedding"],"productionLevels":[5]},{"id":"cohere-embed","name":"Cohere Embed","url":"https://docs.cohere.com/docs/cohere-embed","linkKind":"official-docs","categories":["embedding"]},{"id":"voyage-ai","name":"Voyage AI","url":"https://docs.voyageai.com/docs/embeddings","linkKind":"official-docs","categories":["embedding"],"productionLevels":[5]},{"id":"sentence-transformers","name":"Sentence Transformers","url":"https://github.com/UKPLab/sentence-transformers","linkKind":"github","categories":["embedding"]},{"id":"bge","name":"BGE / FlagEmbedding","url":"https://github.com/FlagOpen/FlagEmbedding","linkKind":"github","categories":["embedding"]},{"id":"vertex-ai-embeddings","name":"Vertex AI Embeddings","url":"https://cloud.google.com/vertex-ai/generative-ai/docs/embeddings/get-text-embeddings","linkKind":"official-docs","categories":["embedding"]},{"id":"azure-openai-embeddings","name":"Azure OpenAI Embeddings","url":"https://learn.microsoft.com/azure/ai-services/openai/concepts/understand-embeddings","linkKind":"official-docs","categories":["embedding"]},{"id":"nomic-embed","name":"Nomic Embed","url":"https://docs.nomic.ai/atlas/embeddings-and-retrieval/text-embedding","linkKind":"official-docs","categories":["embedding"],"productionLevels":[5]},{"id":"firecrawl","name":"Firecrawl","url":"https://github.com/firecrawl/firecrawl","linkKind":"github","categories":["data-extraction"],"productionLevels":[6]},{"id":"scrapy","name":"Scrapy","url":"https://github.com/scrapy/scrapy","linkKind":"github","categories":["data-extraction"],"productionLevels":[6]},{"id":"docling","name":"Docling","url":"https://github.com/docling-project/docling","linkKind":"github","categories":["data-extraction"],"productionLevels":[6]},{"id":"llamaparse","name":"LlamaParse","url":"https://developers.llamaindex.ai/llamaparse/parse/","linkKind":"official-docs","categories":["data-extraction"],"productionLevels":[6]},{"id":"mcp-typescript-sdk","name":"MCP SDK","url":"https://github.com/modelcontextprotocol/typescript-sdk","linkKind":"github","categories":["mcp"]},{"id":"fastmcp","name":"FastMCP","url":"https://github.com/PrefectHQ/fastmcp","linkKind":"github","categories":["mcp"]},{"id":"mcp-registry","name":"MCP Registry","url":"https://github.com/modelcontextprotocol/registry","linkKind":"github","categories":["mcp"]},{"id":"github-mcp-server","name":"GitHub MCP Server","url":"https://github.com/github/github-mcp-server","linkKind":"github","categories":["mcp"]},{"id":"slack-mcp-server","name":"Slack MCP Server","url":"https://github.com/modelcontextprotocol/servers-archived/tree/main/src/slack","linkKind":"github","categories":["mcp"]},{"id":"postgresql-mcp-server","name":"PostgreSQL MCP Server","url":"https://github.com/modelcontextprotocol/servers-archived/tree/main/src/postgres","linkKind":"github","categories":["mcp"]},{"id":"google-drive-mcp-server","name":"Google Drive MCP Server","url":"https://github.com/modelcontextprotocol/servers-archived/tree/main/src/gdrive","linkKind":"github","categories":["mcp"]},{"id":"filesystem-mcp-server","name":"Filesystem MCP Server","url":"https://github.com/modelcontextprotocol/servers/tree/main/src/filesystem","linkKind":"github","categories":["mcp"]},{"id":"nemo-guardrails","name":"NVIDIA NeMo Guardrails","url":"https://github.com/NVIDIA-NeMo/Guardrails","linkKind":"github","categories":["security"]},{"id":"guardrails-ai","name":"Guardrails AI","url":"https://github.com/guardrails-ai/guardrails","linkKind":"github","categories":["security"],"productionLevels":[8]},{"id":"microsoft-presidio","name":"Microsoft Presidio","url":"https://github.com/microsoft/presidio","linkKind":"github","categories":["security"]},{"id":"lakera-guard","name":"Lakera Guard","url":"https://docs.lakera.ai/","linkKind":"official-docs","categories":["security"]},{"id":"prompt-security","name":"Prompt Security","url":"https://prompt.security/","linkKind":"official-product","categories":["security"]},{"id":"protect-ai","name":"Protect AI","url":"https://github.com/protectai","linkKind":"github","categories":["security"]},{"id":"azure-ai-content-safety","name":"Azure AI Content Safety","url":"https://learn.microsoft.com/azure/ai-services/content-safety/overview","linkKind":"official-docs","categories":["security"]},{"id":"aws-bedrock-guardrails","name":"AWS Bedrock Guardrails","url":"https://docs.aws.amazon.com/bedrock/latest/userguide/guardrails.html","linkKind":"official-docs","categories":["security"]},{"id":"langsmith","name":"LangSmith","url":"https://docs.langchain.com/langsmith/observability","linkKind":"official-docs","categories":["observability"],"productionLevels":[1]},{"id":"langfuse","name":"Langfuse","url":"https://github.com/langfuse/langfuse","linkKind":"github","categories":["observability"],"productionLevels":[8]},{"id":"arize-phoenix","name":"Arize Phoenix","url":"https://github.com/Arize-ai/phoenix","linkKind":"github","categories":["observability"],"productionLevels":[1]},{"id":"wandb-weave","name":"Weights & Biases Weave","url":"https://github.com/wandb/weave","linkKind":"github","categories":["observability"]},{"id":"trulens","name":"TruLens","url":"https://github.com/truera/trulens","linkKind":"github","categories":["observability"]},{"id":"ragas","name":"Ragas","url":"https://github.com/explodinggradients/ragas","linkKind":"github","categories":["observability"],"productionLevels":[1]},{"id":"promptfoo","name":"Promptfoo","url":"https://github.com/promptfoo/promptfoo","linkKind":"github","categories":["observability"]},{"id":"helicone","name":"Helicone","url":"https://github.com/Helicone/helicone","linkKind":"github","categories":["observability"],"productionLevels":[8]},{"id":"deepeval","name":"DeepEval","url":"https://github.com/confident-ai/deepeval","linkKind":"github","categories":["observability"],"productionLevels":[1]},{"id":"arize","name":"Arize","url":"https://arize.com/ai-observability-platform/","linkKind":"official-product","categories":["observability"],"productionLevels":[8]},{"id":"mem0","name":"Mem0","url":"https://github.com/mem0ai/mem0","linkKind":"github","categories":["memory"],"productionLevels":[7]},{"id":"zep","name":"Zep","url":"https://github.com/getzep/zep","linkKind":"github","categories":["memory"],"productionLevels":[7]},{"id":"letta","name":"Letta","url":"https://github.com/letta-ai/letta","linkKind":"github","categories":["memory"],"productionLevels":[7]},{"id":"langgraph-memory","name":"LangGraph Memory","url":"https://docs.langchain.com/oss/python/langgraph/add-memory","linkKind":"official-docs","categories":["memory"]},{"id":"redis","name":"Redis","url":"https://github.com/redis/redis","linkKind":"github","categories":["memory","vector-database"]},{"id":"postgresql","name":"PostgreSQL","url":"https://www.postgresql.org/","linkKind":"official-product","categories":["memory"]},{"id":"neo4j","name":"Neo4j","url":"https://neo4j.com/docs/","linkKind":"official-docs","categories":["memory"]},{"id":"chroma","name":"Chroma","url":"https://github.com/chroma-core/chroma","linkKind":"github","categories":["memory","vector-database"],"productionLevels":[4]},{"id":"cognee","name":"Cognee","url":"https://github.com/topoteretes/cognee","linkKind":"github","categories":["memory"],"productionLevels":[7]},{"id":"openai-agents-sdk","name":"OpenAI Agents SDK","url":"https://github.com/openai/openai-agents-python","linkKind":"github","categories":["agents"]},{"id":"pydantic-ai","name":"Pydantic AI","url":"https://github.com/pydantic/pydantic-ai","linkKind":"github","categories":["agents"]},{"id":"semantic-kernel","name":"Semantic Kernel","url":"https://github.com/microsoft/semantic-kernel","linkKind":"github","categories":["agents"]},{"id":"google-adk","name":"Google ADK","url":"https://github.com/google/adk-python","linkKind":"github","categories":["agents"]},{"id":"aws-bedrock-agents","name":"AWS Bedrock Agents","url":"https://docs.aws.amazon.com/bedrock/latest/userguide/agents.html","linkKind":"official-docs","categories":["agents"]},{"id":"azure-ai-agent-service","name":"Microsoft Foundry Agent Service","url":"https://learn.microsoft.com/azure/foundry/agents/overview","linkKind":"official-docs","categories":["agents"]},{"id":"n8n","name":"n8n","url":"https://github.com/n8n-io/n8n","linkKind":"github","categories":["automation"]},{"id":"zapier","name":"Zapier","url":"https://zapier.com/","linkKind":"official-product","categories":["automation"]},{"id":"make","name":"Make","url":"https://www.make.com/","linkKind":"official-product","categories":["automation"]},{"id":"power-automate","name":"Microsoft Power Automate","url":"https://learn.microsoft.com/power-automate/","linkKind":"official-docs","categories":["automation"]},{"id":"temporal","name":"Temporal","url":"https://github.com/temporalio/temporal","linkKind":"github","categories":["automation"]},{"id":"apache-airflow","name":"Apache Airflow","url":"https://github.com/apache/airflow","linkKind":"github","categories":["automation"]},{"id":"prefect","name":"Prefect","url":"https://github.com/PrefectHQ/prefect","linkKind":"github","categories":["automation"]},{"id":"kestra","name":"Kestra","url":"https://github.com/kestra-io/kestra","linkKind":"github","categories":["automation"]},{"id":"pipedream","name":"Pipedream","url":"https://github.com/PipedreamHQ/pipedream","linkKind":"github","categories":["automation"]},{"id":"pinecone","name":"Pinecone","url":"https://docs.pinecone.io/","linkKind":"official-docs","categories":["vector-database"],"productionLevels":[4]},{"id":"weaviate","name":"Weaviate","url":"https://github.com/weaviate/weaviate","linkKind":"github","categories":["vector-database"],"productionLevels":[4]},{"id":"qdrant","name":"Qdrant","url":"https://github.com/qdrant/qdrant","linkKind":"github","categories":["vector-database"]},{"id":"milvus","name":"Milvus","url":"https://github.com/milvus-io/milvus","linkKind":"github","categories":["vector-database"],"productionLevels":[4]},{"id":"pgvector","name":"pgvector","url":"https://github.com/pgvector/pgvector","linkKind":"github","categories":["vector-database"]},{"id":"elasticsearch","name":"Elasticsearch","url":"https://github.com/elastic/elasticsearch","linkKind":"github","categories":["vector-database"]},{"id":"mongodb-atlas-vector-search","name":"MongoDB Atlas Vector Search","url":"https://www.mongodb.com/docs/atlas/atlas-vector-search/","linkKind":"official-docs","categories":["vector-database"]}],"sources":[{"label":"Open-source AI ecosystem breakdown","author":"Rathnakumar Udayakumar","url":"https://www.linkedin.com/posts/rathanuday_the-open-source-ai-ecosystem-is-evolving-activity-7398656721302249472-2C2R","contribution":"Capability taxonomy and the reminder that a production system spans more than model choice."},{"label":"Agentic AI - RAG Tech Stack Layered Blueprint","author":"BhuviAI","url":"https://www.bhuviai.com/agentic-ai-rag-tech-stack-layered-blueprint","contribution":"Production-depth levels and corrections for ambiguous labels in the supplied iceberg image."}]}}
content_markdown: "## Selection rule\n\nThis radar starts from GitHub repositories that appeared in the local Telegram export, then keeps the projects that are useful to AI engineers building real systems: agent runtimes, context pipelines, browser/tool use, generative UI, evaluation, sandboxes, and model infrastructure.\n\nStars and forks were checked from GitHub on 2026-07-23. The ranked ledger is sorted by stars descending, with forks as the secondary signal when popularity is close. \"Export hits\" counts how many link occurrences the repository had in the local Telegram export; public citation stays on the GitHub repository URL.\n\n## How to read the radar\n\nThe human-facing HTML route visualizes the collection as a popularity skyline, an ecosystem map, and five thematic repository lanes. The machine-readable JSON route contains all 140 repository records plus the normalized ecosystem dataset.\n\nProjects are ordered by stars, with forks as the secondary popularity signal. Theme filters do not change that underlying rank.\n\n## Modern AI ecosystem map\n\nThe capability view combines the supplied taxonomy with official product, documentation, and GitHub links. Its 13 lanes cover deployment and inference, LLMs, agentic orchestration, RAG, embeddings, data extraction, MCP, AI security, evaluation and observability, memory, agent SDKs, automation, and vector databases.\n\nThe production-depth view reorganizes a verified subset into nine levels: deployment, evaluation, LLMs, frameworks, vector databases, embeddings, extraction, memory, and alignment. This is deliberately a diagnostic reading aid rather than a claim that every architecture needs one product from every row.\n\nThe supplied iceberg image used ambiguous labels for two framework and extraction entries. The public source text identifies them as LlamaIndex and LlamaParse, so the site uses those verified names instead of reproducing the image labels literally.\n\n## Theme map for the site\n\n### Tools and agent workbenches\n\nOpenClaw, Hermes Agent, Pydantic AI, AgentField, Agent Browser, OpenHuman, Herdr, Worktrunk, InsForge, official skills catalogs, DeerFlow, Cline, Symphony, Superpowers, Claude Code, Codex, Gemini CLI, Warp, Goose, Dify, Agno, Mastra, AutoGen, CrewAI, LangChain, and LangGraph are the operational layer of the radar. They match New Runtime's core observation that AI products are becoming work loops with state, tools, commands, verification, and human approval gates.\n\n### Context engineering and memory\n\nCodeGraph, GitNexus, Supermemory, AgentMemory, Context Mode, LEANN, Claude Context, Agent Reach, Unlimited OCR, Docling, OpenDataLoader PDF, olmOCR, LangExtract, OpenWiki, CocoIndex, Memori, memU, MarkItDown, Firecrawl, Repomix, Cognee, Graphiti, Haystack, MCP servers, FastMCP, and the MCP server catalog are the context supply chain. They make documents, codebases, web pages, tools, and knowledge graphs inspectable by agents instead of dumping raw data into a prompt.\n\n### LLM-UI and generated interfaces\n\nPretext, Impeccable, draw.io MCP, realtime voice components, Taste Skill, design.md, HyperFrames, Stitch Skills, Open Slide, Apps SDK examples, CopilotKit generative UI, json-render, Tambo, and Magentic-UI belong to the generated-interface track. The important trend is not \"AI makes prettier screens\"; it is that the answer can become a task surface with state, validation, user correction, and follow-up actions.\n\n### Evals, specs, and reliability\n\nAutoresearch, Ponytail, Improve, Monty, ZeroLang, ClawKeeper, AutomationBench, MemoryData, FAPO, ALTK Evolve, Agent Skills, SkillSpector, SkillsBench, Evidently, Agent Lightning, system-prompt collections, Spec Kit, OpenAI Evals, DeepEval, OpenEvals, Outlines, A2A, CubeSandbox, and HexStrike are the control plane. They expose whether the agent has a bounded task, structured output, a protocol boundary, a test loop, and an execution sandbox before it touches real systems.\n\n### Model and serving infrastructure\n\nLightpanda, vLLM, Unsloth, and LMCache keep the runtime layer visible. They matter because agent systems are not just prompts; they are compute contracts with browser throughput, serving throughput, memory pressure, routing, fine-tuning, and cost tradeoffs.\n\n## Refresh protocol\n\n1. Re-extract GitHub repository roots from the Telegram export.\n2. Drop placeholders, private repos, inactive redirects that do not resolve, and repos that are not useful to AI engineering.\n3. Refresh GitHub stars, forks, archived state, and canonical full name on the verification date.\n4. Sort the ledger by stars, using forks as the secondary popularity signal.\n5. Re-check the theme labels against New Runtime's site taxonomy before publishing."
featured: true
routes: {"html":"https://newruntime.com/projects/github-projects-for-ai-engineers/","markdown":"https://newruntime.com/projects/github-projects-for-ai-engineers.md","json":"https://newruntime.com/projects/github-projects-for-ai-engineers.json","build_brief":"https://newruntime.com/projects/github-projects-for-ai-engineers/build-brief.md","acceptance":"https://newruntime.com/projects/github-projects-for-ai-engineers/acceptance.md"}
source_format: "markdown"
---

# GitHub Projects for AI Engineers

## Retrieval answer

A Telegram-corpus-derived GitHub radar plus a source-backed map of the models, orchestration, retrieval, memory, security, automation, and infrastructure layers around production AI systems. Turn the Telegram export and two public ecosystem references into a ranked repository radar and a navigable map of the wider production AI stack.

## Selection rule

This radar starts from GitHub repositories that appeared in the local Telegram export, then keeps the projects that are useful to AI engineers building real systems: agent runtimes, context pipelines, browser/tool use, generative UI, evaluation, sandboxes, and model infrastructure.

Stars and forks were checked from GitHub on 2026-07-23. The ranked ledger is sorted by stars descending, with forks as the secondary signal when popularity is close. "Export hits" counts how many link occurrences the repository had in the local Telegram export; public citation stays on the GitHub repository URL.

## How to read the radar

The human-facing HTML route visualizes the collection as a popularity skyline, an ecosystem map, and five thematic repository lanes. The machine-readable JSON route contains all 140 repository records plus the normalized ecosystem dataset.

Projects are ordered by stars, with forks as the secondary popularity signal. Theme filters do not change that underlying rank.

## Modern AI ecosystem map

The capability view combines the supplied taxonomy with official product, documentation, and GitHub links. Its 13 lanes cover deployment and inference, LLMs, agentic orchestration, RAG, embeddings, data extraction, MCP, AI security, evaluation and observability, memory, agent SDKs, automation, and vector databases.

The production-depth view reorganizes a verified subset into nine levels: deployment, evaluation, LLMs, frameworks, vector databases, embeddings, extraction, memory, and alignment. This is deliberately a diagnostic reading aid rather than a claim that every architecture needs one product from every row.

The supplied iceberg image used ambiguous labels for two framework and extraction entries. The public source text identifies them as LlamaIndex and LlamaParse, so the site uses those verified names instead of reproducing the image labels literally.

## Theme map for the site

### Tools and agent workbenches

OpenClaw, Hermes Agent, Pydantic AI, AgentField, Agent Browser, OpenHuman, Herdr, Worktrunk, InsForge, official skills catalogs, DeerFlow, Cline, Symphony, Superpowers, Claude Code, Codex, Gemini CLI, Warp, Goose, Dify, Agno, Mastra, AutoGen, CrewAI, LangChain, and LangGraph are the operational layer of the radar. They match New Runtime's core observation that AI products are becoming work loops with state, tools, commands, verification, and human approval gates.

### Context engineering and memory

CodeGraph, GitNexus, Supermemory, AgentMemory, Context Mode, LEANN, Claude Context, Agent Reach, Unlimited OCR, Docling, OpenDataLoader PDF, olmOCR, LangExtract, OpenWiki, CocoIndex, Memori, memU, MarkItDown, Firecrawl, Repomix, Cognee, Graphiti, Haystack, MCP servers, FastMCP, and the MCP server catalog are the context supply chain. They make documents, codebases, web pages, tools, and knowledge graphs inspectable by agents instead of dumping raw data into a prompt.

### LLM-UI and generated interfaces

Pretext, Impeccable, draw.io MCP, realtime voice components, Taste Skill, design.md, HyperFrames, Stitch Skills, Open Slide, Apps SDK examples, CopilotKit generative UI, json-render, Tambo, and Magentic-UI belong to the generated-interface track. The important trend is not "AI makes prettier screens"; it is that the answer can become a task surface with state, validation, user correction, and follow-up actions.

### Evals, specs, and reliability

Autoresearch, Ponytail, Improve, Monty, ZeroLang, ClawKeeper, AutomationBench, MemoryData, FAPO, ALTK Evolve, Agent Skills, SkillSpector, SkillsBench, Evidently, Agent Lightning, system-prompt collections, Spec Kit, OpenAI Evals, DeepEval, OpenEvals, Outlines, A2A, CubeSandbox, and HexStrike are the control plane. They expose whether the agent has a bounded task, structured output, a protocol boundary, a test loop, and an execution sandbox before it touches real systems.

### Model and serving infrastructure

Lightpanda, vLLM, Unsloth, and LMCache keep the runtime layer visible. They matter because agent systems are not just prompts; they are compute contracts with browser throughput, serving throughput, memory pressure, routing, fine-tuning, and cost tradeoffs.

## Refresh protocol

1. Re-extract GitHub repository roots from the Telegram export.
2. Drop placeholders, private repos, inactive redirects that do not resolve, and repos that are not useful to AI engineering.
3. Refresh GitHub stars, forks, archived state, and canonical full name on the verification date.
4. Sort the ledger by stars, using forks as the secondary popularity signal.
5. Re-check the theme labels against New Runtime's site taxonomy before publishing.