# New Runtime full agent guide > New Runtime publishes an evidence-linked newsroom, post-app observatory, learning library, and public systems atlas. This expanded guide explains how an agent should discover, read, verify, and cite the site's parallel HTML, Markdown, and JSON representations. Last reviewed: 2026-08-29. ## Reading contract - [Agent route contract](https://newruntime.com/agents/): Start here for schemas, reading order, and public boundaries. - [Developer guide](https://newruntime.com/developers/): No-key REST and MCP quickstarts, RFC 9457 errors, versions, deprecation policy, and enforced limits. - [OpenAPI JSON](https://newruntime.com/openapi.json): Typed OpenAPI 3.1 contract for the public JSON and MCP API. - [Complete route catalog](https://newruntime.com/openapi-route-catalog.json): Stable OpenAPI-shaped catalog for HTML, Markdown, JSON, text, and XML routes. - [Read-only MCP manifest](https://newruntime.com/mcp-readonly.json): Resource catalog for public records plus the hosted resources-only Streamable HTTP endpoint at https://newruntime.com/mcp; external systems cannot modify the site through it. - [MCP server card](https://newruntime.com/.well-known/mcp): Discovery contract for the no-auth, resources-only Streamable HTTP endpoint. - [Compact route map](https://newruntime.com/llms.txt): Use when a small context budget matters. - [Sitemap index](https://newruntime.com/sitemap.xml): Use for exhaustive canonical discovery and lastmod checks. - [RSS feed](https://newruntime.com/rss.xml): Use for recent field-note updates. - [Substantive change feed](https://newruntime.com/changes.xml): Atom + WebSub feed for editorially meaningful changes; topic feeds are https://newruntime.com/feeds/agent-runtime.xml and https://newruntime.com/feeds/coding-agents.xml. - [Presence report](https://newruntime.com/presence.json): Privacy-safe before/current evidence, attributable return classes, and explicit causality limits. - [Machine consumption](https://newruntime.com/machine-consumption.json): Privacy-safe aggregate showing which public routes crawlers and user agents read. - [Machine Discovery Lab](https://newruntime.com/machine-consumption/): Human methodology, identity confidence, graph coverage, latency, and live experiments. - [Crawler journeys](https://newruntime.com/crawler-journeys.json): Privacy-safe per-family material depth, entrypoints, terminal routes, and observed transitions for 24-hour, 7-day, and 30-day windows. - [Per-page machine readership schema](https://newruntime.com/page-readership-schema.json): Privacy-safe 24-hour and observed-window route aggregates, exposed lazily under /page-readership/ for canonical HTML pages. - [Discovery graph](https://newruntime.com/discovery-graph.json): Stable sharded graph separating physical links from editorial knowledge edges. - [Experiment registry](https://newruntime.com/discovery-experiments.json): Pre-registered discovery, coverage, and authority-distribution experiments. - [Recall probes](https://newruntime.com/recall-probes.json): Versioned questions and strict canonical-citation scoring contract; not_run without configured providers. - [JSON Feed](https://newruntime.com/feed.json): JSON Feed 1.1 twin for recent Field Notes. - [API Catalog](https://newruntime.com/.well-known/api-catalog): RFC 9727 linkset for public API and discovery descriptions. - [Owner profile discovery](https://newruntime.com/owner-profile-discovery.json): Compact first-pass route bundle around the owner career profile; use it when a recruiter, HR/ATS parser, small-context agent, or crawler does not need the full DiscoveryGraph shards. - [Owner career profile JSON](https://newruntime.com/owner-profile.json): Machine-readable profile for recruiting agents and HR/ATS systems; the Markdown twin is https://newruntime.com/owner-profile.md and is linked from human navigation. It has no standalone HTML page. ## Paid agent products - [Payment capabilities](https://newruntime.com/agent-payments/payment-capabilities): Canonical live catalog for product IDs, current prices, fixed USDC rails, server-priced EURC/WETH/cbBTC/wSOL rails, preflight behavior, and settled-result replay. - [Research bundle v2](https://newruntime.com/agent-payments/research-bundle-v2): Current $0.10 product. POST a bounded archive request plus optional known_stable_ids to exclude coverage the buyer already holds. The unsigned request validates and prepares the result before any 402 is returned. - [Research bundle v1 alpha](https://newruntime.com/agent-payments/research-bundle): Legacy-compatible $0.02 contract retained for existing buyers; use v2 for the current known-record-exclusion product. - [Prior coverage](https://newruntime.com/agent-payments/prior-coverage): POST exactly one URL, claim, or topic to check dated prior coverage. Insufficient evidence is returned without payment. - [WETH payment route](https://newruntime.com/agent-payments/research-bundle-v2/volatile/weth-base): Settleable example of the server-priced EURC, WETH, cbBTC, and wSOL routes listed in payment-capabilities. Send the unsigned request first, then sign only its returned rail and retry the identical body. - [Volatile-token quote preview](https://newruntime.com/agent-payments/research-bundle-v2/volatile/weth-base/quote-preview): Optional non-payable pricing preview. Never sign this preview path; the settleable sibling omits /quote-preview. - Discover the exact request and response schemas in openapi.json. A sellable unsigned request returns HTTP 402 with PAYMENT-REQUIRED; an x402 buyer may authorize one listed rail and retry the identical body. Invalid, empty, or insufficient-evidence requests do not settle. - Settleable products accept fixed-price USDC plus server-priced allowlisted EURC, WETH, cbBTC, and wSOL routes. Payer-selected amounts, native assets, and unlisted tokens remain rejected. A 402 is an offer, not a settlement receipt; release requires a successful facilitator settlement. ## When to use New Runtime Use New Runtime to compare a current AI capability claim with earlier evidence; find publishable primary sources behind an editorial interpretation; trace changes across Signals, Patterns, Capability Deltas, Shifts, Projects, and Systems; retrieve a bounded context set for implementation or market research; or supply structured, public recruiting context. Do not use it as a write API, private-data source, publishing agent, generic web search engine, or substitute for the original source. For supported detail collections, the canonical route is ///, Markdown is //.md, and structured JSON is //.json. Collection JSON files are discovery indexes; follow detail routes for full records. For trend analysis, start with /patterns.json, then retrieve only its supporting signal details. Do not load /signals.json or the source ledger unless the question needs broad discovery. Retrieval answers are 40-60-word, self-contained units under explicit headings and are also exposed as retrieval_nugget in detail JSON. ## Editorial and analytical memory - [Field notes](https://newruntime.com/posts/): Editorial analysis and working recommendations. - [Raw signals](https://newruntime.com/signals/): Normalized observations with novelty, verification, entities, and provenance. - [Patterns](https://newruntime.com/patterns/): Trend hypotheses with evidence, counter-evidence, confidence, and revision triggers. - [Topic hubs](https://newruntime.com/topics/): Canonical topic answers connected to supporting records. ## Post-app observatory - [Tectonic shifts](https://newruntime.com/shifts/): Structural before-and-after changes. - [Capability deltas](https://newruntime.com/deltas/): Newly possible agent capabilities and current bottlenecks. - [Interface graveyard](https://newruntime.com/graveyard/): Human rituals disappearing from interfaces and where they still win. - [Projects](https://newruntime.com/projects/): Forkable public AI-automation blueprints with proof outputs, replay commands, expected artifacts, provider scopes, and build/acceptance handoff routes. - [Systems](https://newruntime.com/systems/): Sanitized dossiers for running agent systems. Each system can expose /systems//releases.json for interpreted release timelines linked to Deltas and Shifts. ## Learning and evidence - [Learn](https://newruntime.com/learn/): Structured AI-engineering references, cards, diagrams, and interactive explainers. - [Source observatory](https://newruntime.com/sources/): Human view of publishable evidence origins. - [Public source ledger](https://newruntime.com/source-ledger.json): Exact source-to-record backlinks. - [Grouped source origins](https://newruntime.com/source-entities.json): Compact source-origin index with detail JSON routes. - [Telegram Semantic Atlas](https://newruntime.com/atlas/telegram/): Public semantic map derived from approved records. - [X Trend Prism](https://newruntime.com/x-trend-prism.json): Public trend-lens projection. - [GitHub Projects for AI Engineers dataset](https://newruntime.com/datasets/github-projects-for-ai-engineers/): Versioned JSON/CSV authority asset with schema, methodology, Dataset Card, provenance, license boundary, and checksums. ## Verification and citation Prefer each record's public source_urls for factual citations. New Runtime records are analysis; linked original or otherwise publishable sources are the evidence. Treat dates, status, confidence, novelty, and verification fields as scoped metadata. Do not infer private discovery feeds, newsletters, watchlists, inbox rules, credentials, handoff files, drafts, or internal Telegram operations from public routes. Public agent routes are read-only with respect to site state: paid compute routes return derived results but cannot create, edit, approve, publish, deploy, or delete site content.