{ "schema_version": "newruntime-topic-hub-v0.1", "type": "topic_hub", "slug": "agent-runtime", "title": "Agent runtime - New Runtime", "description": "The runtime layer around agents: sandboxes, tools, hosted execution, permissions, state, and deployment surfaces.", "answer": [ "The agent runtime is the system that lets models safely use tools, files, browsers, terminals, and external services.", "As agents take real actions, runtime concerns become product concerns: isolation, permissions, logging, replay, and rollback.", "This is where demos turn into repeatable infrastructure." ], "search_intents": [ "agent runtime", "AI agent infrastructure", "sandboxed agent execution" ], "status": "featured", "last_updated": "2026-07-22", "counts": { "total": 11, "signals": 9, "patterns": 1, "posts": 1, "atlas": 0, "sources": 23 }, "routes": { "html": "https://newruntime.com/topics/agent-runtime/", "markdown": "https://newruntime.com/topics/agent-runtime.md", "json": "https://newruntime.com/topics/agent-runtime.json" }, "top_sources": [ "https://agentos-sdk.dev/", "https://claude.com/blog/new-in-claude-managed-agents", "https://cline.bot/blog/introducing-cline-sdk-the-upgraded-agent-runtime", "https://developers.cloudflare.com/sandbox/tutorials/devin-outposts/", "https://docs.inference.net/introduction", "https://github.com/anthropics/claude-cookbooks/blob/main/claude_agent_sdk/07_Hosting_the_agent.ipynb", "https://github.com/bytedance/deer-flow", "https://github.com/cline/cline/tree/main/sdk/apps/examples", "https://github.com/langchain-ai/deepagents", "https://github.com/NousResearch/hermes-agent", "https://github.com/rivet-dev/agentos", "https://langchain.com/blog/give-your-ai-agent-its-own-computer" ], "patterns": [ { "kind": "Pattern", "slug": "harness-architecture-outlives-model-choice", "title": "Harness architecture outlives model choice", "description": "For production agents, the harness is becoming a more durable product boundary than the identity of the model running inside it.", "date": "2026-07-16", "topics": [ "harness-engineering", "agent-runtime", "model-routing" ], "source_count": 5, "metric": "high", "url": "https://newruntime.com/patterns/harness-architecture-outlives-model-choice/" } ], "field_notes": [ { "kind": "Field Note", "slug": "devin-outposts-agent-execution-plane", "title": "Devin Outposts Splits the Agent Brain from the Execution Plane", "description": "Devin Outposts moves command execution, repository access, and sandbox lifecycle into customer-controlled infrastructure while the agent loop remains in Cognition's cloud.", "date": "2026-07-22", "topics": [ "coding-agents", "agent-runtime", "infrastructure" ], "source_count": 4, "url": "https://newruntime.com/posts/devin-outposts-agent-execution-plane/" } ], "raw_signals": [ { "kind": "Raw Signal", "slug": "render-packages-deployment-primitives-for-agent-workloads", "title": "Render Packages Deployment Primitives for Agent Workloads", "description": "Render combines services, workers, databases, cron jobs, previews, and infrastructure definitions into a deployment surface suitable for long-running agent systems.", "date": "2026-07-11", "topics": [ "deployment", "agent-runtime", "infrastructure" ], "source_count": 3, "metric": "notable", "url": "https://newruntime.com/signals/render-packages-deployment-primitives-for-agent-workloads/" }, { "kind": "Raw Signal", "slug": "agentos-lightweight-runtime-for-coding-agents", "title": "agentOS packages a lightweight runtime for coding agents", "description": "Rivet's agentOS combines isolated execution, durable workflows, resource limits, and host-side secrets in a WebAssembly and Rust runtime.", "date": "2026-06-29", "topics": [ "agent-runtime", "sandboxes", "coding-agents" ], "source_count": 3, "metric": "notable", "url": "https://newruntime.com/signals/agentos-lightweight-runtime-for-coding-agents/" }, { "kind": "Raw Signal", "slug": "deerflow-2-0-packages-an-open-runtime-for-autonomous-agents", "title": "DeerFlow 2.0 Packages an Open Runtime for Autonomous Agents", "description": "ByteDance's open runtime combines planning, subagents, files, tools, and long-running execution into a deployable system.", "date": "2026-06-28", "topics": [ "open-source", "multi-agent", "agent-runtime" ], "source_count": 1, "metric": "notable", "url": "https://newruntime.com/signals/deerflow-2-0-packages-an-open-runtime-for-autonomous-agents/" }, { "kind": "Raw Signal", "slug": "ai-sdk-7-grows-into-an-agent-runtime", "title": "AI SDK 7 Grows into an Agent Runtime", "description": "Vercel expands a model integration library with tool loops, state, transport, and UI primitives needed for production agent applications.", "date": "2026-06-27", "topics": [ "ai-sdk", "agent-runtime", "vercel" ], "source_count": 1, "metric": "notable", "url": "https://newruntime.com/signals/ai-sdk-7-grows-into-an-agent-runtime/" }, { "kind": "Raw Signal", "slug": "the-computer-becomes-a-runtime-assigned-to-the-agent", "title": "The Computer Becomes a Runtime Assigned to the Agent", "description": "Giving each agent an isolated machine turns browser, shell, files, and installed software into controlled execution infrastructure.", "date": "2026-06-19", "topics": [ "sandboxes", "agent-runtime", "computer-use" ], "source_count": 1, "metric": "notable", "url": "https://newruntime.com/signals/the-computer-becomes-a-runtime-assigned-to-the-agent/" }, { "kind": "Raw Signal", "slug": "tokens-stay-the-cost-while-runtime-becomes-the-product", "title": "Tokens Stay the Cost While Runtime Becomes the Product", "description": "The differentiated layer shifts toward execution, tools, policy, state, and user workflow while model inference remains metered input.", "date": "2026-06-18", "topics": [ "agent-runtime", "product-strategy", "ai-economics" ], "source_count": 1, "metric": "notable", "url": "https://newruntime.com/signals/tokens-stay-the-cost-while-runtime-becomes-the-product/" }, { "kind": "Raw Signal", "slug": "the-agent-becomes-a-separate-hosting-layer", "title": "The Agent Becomes a Separate Hosting Layer", "description": "Anthropic's Agent SDK hosting pattern separates model calls from a persistent process that owns tools, state, execution, and recovery.", "date": "2026-06-16", "topics": [ "agent-runtime", "hosting", "agent-sdk" ], "source_count": 2, "metric": "notable", "url": "https://newruntime.com/signals/the-agent-becomes-a-separate-hosting-layer/" }, { "kind": "Raw Signal", "slug": "cline-sdk-exposes-an-open-harness-for-coding-agents", "title": "Cline SDK Exposes an Open Harness for Coding Agents", "description": "Cline separates the reusable agent runtime from its interface, making tool execution and coding workflows embeddable in other products.", "date": "2026-05-14", "topics": [ "coding-agents", "sdk", "agent-runtime" ], "source_count": 2, "metric": "notable", "url": "https://newruntime.com/signals/cline-sdk-exposes-an-open-harness-for-coding-agents/" }, { "kind": "Raw Signal", "slug": "claude-managed-agents-add-outcomes-and-multi-agent-orchestration", "title": "Claude Managed Agents Add Outcomes and Multi-Agent Orchestration", "description": "Anthropic's managed-agent layer adds long-running outcomes, orchestration, and background work around the model call.", "date": "2026-05-07", "topics": [ "managed-agents", "orchestration", "agent-runtime" ], "source_count": 1, "metric": "notable", "url": "https://newruntime.com/signals/claude-managed-agents-add-outcomes-and-multi-agent-orchestration/" } ], "atlas_records": [] }