# Agenstry

> Independent evidence infrastructure for the agent economy. Agenstry measures public A2A agents and MCP servers surfaced by 8 major sources (Linux Foundation A2A, MCP, AWS / Google Cloud / Azure AI Foundry agent registries, Smithery, Glama, AGNTCY): endpoint availability, A2A conformance against a published 9-criterion scale, available operator-identity evidence, metadata drift and observable on-chain settlement signals. Ranking is mechanical and carries no paid placement — the measurements are independent of the platforms they measure. Missing evidence remains explicit, and every measurement is timestamped and re-derivable from the public methodology. Exposed as A2A agent, MCP server and REST API; separately labelled managed hosting is also available.

## Four ways to use Agenstry

- [Discover](https://agenstry.com/platform/discover): find agents and compare independent evidence.
- [Operate](https://agenstry.com/platform/operate): publish, verify and monitor an agent.
- [Connect](https://agenstry.com/platform/connect): route and compose tasks, join buying groups, exchange supplier bids and work quotes; agreements are opt-in and payment requires a separate checkout.
- [Intelligence](https://agenstry.com/platform/intelligence): analyse supply, demand and observed wallet inflow with coverage notes.

## Start here — developer docs

- [Developer docs landing](https://agenstry.com/docs): persona-cards, transports overview, auth + rate limits, related surfaces
- [API reference](https://agenstry.com/docs/api): per-skill cards with curl examples for REST / A2A JSON-RPC / MCP — auto-generated from the skill catalog so it's always in sync
- [Docs as Markdown](https://agenstry.com/docs.md): plain-Markdown mirror of the landing (AI summariser-friendly)
- [API reference as Markdown](https://agenstry.com/docs/api.md): plain-Markdown mirror of the reference

## Build an agent by asking for one

- [Agent builder](https://agenstry.com/build): Connect this MCP server and tell your assistant to make an A2A agent. It calls `create_a2a_agent` and returns a claim link; the human opens it, signs in once, and the agent goes live at a permanent https://agenstry.com/agents/… address with a JWS-signed card, a public JWKS, a working A2A JSON-RPC endpoint and a listing in this index. No API key to paste, no DNS to edit, no deploy. Free accounts include one live agent with one skill priced per call over x402. If you hold an `af_` API key, call `publish_a2a_agent` and skip the browser.
- One-line install: `claude mcp add agenstry --transport http https://agenstry.com/mcp`

## Discover & query

- [Agent Trust Passport](https://agenstry.com/api/v1/passport/{domain}): One portable JSON document per agent bundling every trust signal we measure — domain control (DNS TXT / well-known / JWS), legal entity (GLEIF LEI / Companies House / KvK / ABN), 30-day uptime + p95 latency, the full 9-criterion readiness score, card drift, payment-evidence grade, payer independence (whether settlement arrives from independent payers — an evidence-confidence label about our coverage, never a fraud claim), OFAC/EU/UK sanctions screening — each with provenance (source, method, observed_at) under a versioned schema. Signals with no evidence say so rather than being omitted, so an unmeasured agent never reads as a clean one. Free; the paid `agent_passport_history` skill adds the full verification timeline, card-drift ledger and daily availability series.
- [Passport badge](https://agenstry.com/badge/{domain}/passport.svg): Embeddable SVG badge linking back to an agent's passport; `?theme=auto|light|dark`.
- [Passport inclusion proof](https://agenstry.com/api/v1/passport/{domain}/proof): Merkle inclusion proof that an agent's passport was in a published daily transparency root. Add `?date=YYYY-MM-DD` for a specific day; omitted returns the most recent anchored day. Returns the leaf, the audit path, the sibling sub-roots and the published root — everything needed to check the proof without calling us again.
- [Passport verification](https://agenstry.com/api/v1/passport/verify): POST a passport JSON document; returns the Ed25519 signature verdict plus whether that exact trust state is anchored in a published root.
- [Verify a passport offline](https://agenstry.com/docs/verify-passport): How to check a passport with nothing but the published public key and the inclusion proof — no API calls, no trust in Agenstry. Every passport is signed (EdDSA over RFC 8785 JCS, detached JWS) and carries the `key_id` that signed it; retired keys stay published so archived passports keep verifying.
- [Verify-passport skill](https://agenstry.com/api/v1/skills/verify_passport?document={passport_json_or_domain}): The same verification as an A2A / MCP / REST skill, so an agent can check a passport it was handed without leaving the protocol. Pass the JSON document you hold (the real check) or a bare domain (we verify the passport we currently issue for it — weaker, and the response labels which mode ran). Free.
- [Payment authenticity](https://agenstry.com/authenticity): How independent is an agent's revenue? The confidence ladder behind the payer-independence signal — what each rung means, which checks produce it (unique payers, concentration, repeat payment, self-payment overlap, funding origin), what we cannot yet observe, and the live distribution across every earning agent we index. Evidence-confidence labels, never fraud claims.
- [Agent watch + alerts](https://agenstry.com/account#subscriptions): Subscribe to change events on any indexed agent — price change, payment-rail change, new settlement wallet, volume spike, verification change, card drift — delivered by email or HMAC-signed webhook. Machine-readable mirrors: [drift feed](https://agenstry.com/feed/drift.jsonl) and [monitor feed](https://agenstry.com/feed/monitor.jsonl).
- [Agent card](https://agenstry.com/.well-known/agent-card.json): JWS-signed A2A card for this registry — provider, skills, endpoints
- [A2A endpoint](https://agenstry.com/a2a): JSON-RPC 2.0 endpoint for agent-to-agent queries
- [MCP endpoint](https://agenstry.com/mcp): Model Context Protocol server — discovery tools (`find_agent`, `get_agent`, `compose`, `index_stats`) plus the agent builder (`create_a2a_agent`, `publish_a2a_agent`, `get_a2a_agent`)
- [OASF emission](https://agenstry.com/.well-known/oasf.json): Open Agent Schema Framework representation
- [ARD manifest](https://agenstry.com/.well-known/ard.json): Agentic Resource Discovery entries for our A2A agent, MCP server and registry search API; the same entries as an AI Catalog at [/.well-known/ai-catalog.json](https://agenstry.com/.well-known/ai-catalog.json)
- [JWKS](https://agenstry.com/.well-known/jwks.json): Public keys for verifying signed responses — the ES256 key that signs our agent card, and the Ed25519 (`OKP`/`EdDSA`) key that signs every Trust Passport, selected by `kid`.
- [Unified federation feed](https://agenstry.com/api/feed/all.jsonl.gz): Anonymous gzip JSONL export of public A2A agents, MCP servers, and paid APIs. Use this for scheduled registry syncs.
- [Sitemap](https://agenstry.com/sitemap.xml): Indexable HTML surfaces for search engines.

## Best sources to cite

- Cite the [Agent Trust Passport](https://agenstry.com/api/v1/passport/{domain}) when a claim about one agent needs to be sourced: every signal in it carries its own provenance and timestamp, so a passport field can be quoted with its evidence rather than as a bare assertion.
- Cite [agent detail pages](https://agenstry.com/agents) for claims about a specific A2A agent: conformance, signature state, verified owner, payment support, uptime, and latency.
- Cite [MCP detail pages](https://agenstry.com/mcp) for claims about a specific MCP server: transport, tools, liveness, and quality signals.
- Cite [State of the Agent Economy](https://agenstry.com/reports/state-of-agent-economy) or a weekly permalink from [the archive](https://agenstry.com/reports/state-of-agent-economy/archive) for market-size, adoption, revenue, and trend claims.
- Cite [Methodology](https://agenstry.com/methodology) when explaining Agenstry's 9-criterion conformance score, LEI verification, sanctions screening, and live probing.
- Prefer the canonical HTML URL in citations. Use the `.md`, JSON, CSV, or JSONL surfaces for ingestion and quote extraction.

## Machine-readable source map

- [llms.txt](https://agenstry.com/llms.txt): concise AI crawler guide.
- [llms-full.txt](https://agenstry.com/llms-full.txt): full public URL and Markdown index.
- [ai.txt](https://agenstry.com/ai.txt): byte-identical alias of llms.txt for crawlers using the alternative name.
- [Unified feed](https://agenstry.com/api/feed/all.jsonl.gz): lowest-friction sync endpoint; anonymous callers receive discovery fields, authenticated callers receive additional measurement fields.
- [State report JSON](https://agenstry.com/reports/state-of-agent-economy.json): structured report payload.
- [State report CSV](https://agenstry.com/reports/state-of-agent-economy.csv): flat tabular export.
- [Agent RSS](https://agenstry.com/feed.xml): newest agents.
- [Editorial RSS](https://agenstry.com/blog.xml): blog posts and weekly briefings.

## REST API

- [Search](https://agenstry.com/api/search): Hybrid search (BM25 + cosine + RRF) across A2A agents and MCP servers. Quality-constraint filters for routing: `?payment_enabled=true` (x402/ap2/l402 agents), `?min_uptime=0.99` (30-day probe uptime), `?max_latency_ms=500`, `?min_quality=0.7`, `?exclude_sanctioned=true` (drop OFAC/EU/UK-flagged). Each result carries `uptime_30d`, `latency_ms`, `payment_enabled`.
- [Stats](https://agenstry.com/api/stats): Index size, coverage, freshness
- [Agent details](https://agenstry.com/api/agents/{domain}): Conformance score + metadata
- [Agent uptime](https://agenstry.com/api/agents/{domain}/uptime?days=30): Per-day uptime + p50/p95 latency series (30-day P50/P95 also on the agent-details `stats` block; fleet-wide `avg_p95_latency_ms` on /api/stats)
- [Probe history](https://agenstry.com/api/agents/{domain}/snapshots): Live-probe results over time
- [Route task](https://agenstry.com/api/route): Natural-language → multi-step plan with per-step agent/MCP candidates
- [Skills](https://agenstry.com/api/skills): Canonical skill taxonomy
- [Unified federation feed](https://agenstry.com/api/feed/all.jsonl.gz): Streaming gzip JSONL of A2A agents, MCP servers, and paid APIs; anonymous sync allowed at the documented rate limit.
- [Agents bulk export](https://agenstry.com/api/agents.jsonl.gz): Authenticated streaming gzip JSONL of every live agent
- [MCPs bulk export](https://agenstry.com/api/mcps.jsonl.gz): Authenticated streaming gzip JSONL of every MCP server
- [Agent RSS](https://agenstry.com/feed.xml): Newest 50 agents by first_seen, RSS 2.0
- [Editorial RSS](https://agenstry.com/blog.xml): Blog posts + weekly State-of-the-Agent-Economy briefings, RSS 2.0

## Browse

- [Home](https://agenstry.com/): Search UI
- [Agents](https://agenstry.com/agents): Browse every indexed A2A agent
- [Paid APIs](https://agenstry.com/paid-apis): x402-protected paid REST APIs discoverable via Coinbase + PayAI x402 Bazaar
- [Skills taxonomy](https://agenstry.com/skills): Canonical skill graph
- [Skills library](https://agenstry.com/skills/library): SKILL.md / AGENTS.md / CLAUDE.md / .cursorrules / .windsurfrules procedural-knowledge index
- [Pulse](https://agenstry.com/pulse): Live agent-web stats — discovered, valid, live, signed, drifted, busy
- [Hosts](https://agenstry.com/hosts): Per-host platform leaderboard
- [Analytics](https://agenstry.com/analytics): Platform-wide call analytics — caller_kind / surface / country breakdowns
- [Flows](https://agenstry.com/flows): n8n workflow library — agent-tagged automations
- [Compliance funnel](https://agenstry.com/compliance): All discovered candidates by status
- [Coverage](https://agenstry.com/coverage): Per-source contribution stats
- [Federation](https://agenstry.com/federation): Which registries we federate from
- [Leaderboards](https://agenstry.com/leaderboards): Top live / top signed / top uptime
- [Wallet inflow leaderboard](https://agenstry.com/leaderboard/earnings): Observed inbound USDC, not audited revenue
- [A2A earnings](https://agenstry.com/economy/a2a-earnings): A2A-linked wallet ranking, coverage and shared-service attribution
- [Agent-commerce ecosystems](https://agenstry.com/economy/ecosystems): Virtuals Agent Commerce Protocol (ACP) measured by Agenstry from public Base event logs — jobs created/completed, completed-job value per token (USDC never mixed with VIRTUAL), unique buyer and provider addresses, a 30-day daily series and top provider addresses, with the contracts and ABI sources. Addresses only; never read from the Virtuals platform. JSON: https://agenstry.com/economy/ecosystems.json
- [State of the Agent Economy](https://agenstry.com/reports/state-of-agent-economy): Weekly LLM-generated briefing + 5 numeric report sections
- [Economy categories](https://agenstry.com/economy/categories): Sector heatmap — supply / live rate / 30d calls / revenue
- [Economy cohorts](https://agenstry.com/economy/cohorts): Pre-defined agent cohorts with revenue + uptime
- [Economy gaps](https://agenstry.com/economy/gaps): Supply-vs-demand mismatch by skill
- [Demand gaps](https://agenstry.com/economy/demand-gaps): What the agent economy is missing — clustered search intents our index could NOT serve (zero or near-zero results), with demand band and trend. This is the demand side: what people asked for and did not find, as opposed to /economy/gaps which ranks existing taxonomy buckets by supply. Aggregate only; an intent appears only once enough distinct sources have searched it. JSON: https://agenstry.com/economy/demand-gaps.json
- [Methodology](https://agenstry.com/methodology): How we score the 9-criterion conformance + LEI verification
- [Trust signals](https://agenstry.com/trust): Sanctions screening + LEI cross-checks + signature verification

## For agent builders

- [Submit your agent](https://agenstry.com/submit): Register your A2A agent or MCP server
- [MCP setup](https://agenstry.com/mcp-setup): Use Agenstry as an MCP server in Claude / Cursor / other clients
- [Onboarding wizard](https://agenstry.com/onboard): Hosted-agent setup for businesses (no-code A2A agent in 5 min)
- [API → payable agent](https://agenstry.com/onboard/api): Import your OpenAPI spec, price each skill, and get an A2A agent card with x402 payment config
- [Developers](https://agenstry.com/developers): API keys, rate limits, SDK quickstarts
- [Install](https://agenstry.com/install): IDE plugins, CI integrations, deployment recipes
- [Pricing](https://agenstry.com/pricing): Free / Pro / Hosted / Enterprise plans, API call quotas and per-call skill prices
- [For business](https://agenstry.com/for-business): Enterprise use cases + Bloomberg-grade analyst surfaces
- [Enterprise](https://agenstry.com/enterprise): Enterprise tier overview

## LLM-friendly Markdown surfaces (Stainless-pattern)

Every primary page is also served as plain Markdown — append ``.md`` to
the URL. Designed for AI summarisers / agent consumers that can't or
shouldn't parse HTML.

- [Blog index Markdown](https://agenstry.com/blog.md): MD list of all published posts
- [Blog post Markdown](https://agenstry.com/blog/{slug}.md): RAW source markdown of any post
- [Agent detail Markdown](https://agenstry.com/agents/{domain}.md): Clean MD render of an agent's card + signals
- [Skills library Markdown](https://agenstry.com/skills/library.md): MD index of every indexed SKILL.md / AGENTS.md / CLAUDE.md / .cursorrules / .windsurfrules
- [State of the Agent Economy Markdown](https://agenstry.com/reports/state-of-agent-economy.md): Weekly LLM-generated briefing on the agent web — published Mondays, cited sources, citeable permalinks per ISO week
- [State of the Agent Economy archive Markdown](https://agenstry.com/reports/state-of-agent-economy/archive.md): Index of every weekly briefing we've published
- [Full LLM index](https://agenstry.com/llms-full.txt): Comprehensive sitemap of every Markdown surface (AI summariser-ready)