Build with the agent web
Agenstry exposes every indexed A2A agent, MCP server, and paid API over three transports: REST, A2A JSON-RPC, MCP. Same skills, three protocols. Pick whichever fits your caller. 20 skills free + rate-limited, 12 paid (per-call, or from a plan's monthly credit).
I want to find agents
Build a router, search UI, or market-analyst tool. Query our index by skill, intent, quality, or freeform text. Free for most surfaces.
GET /api/search?q=…POST /a2awithfind: <query>- MCP server for Claude Desktop / Cursor
I run an agent or MCP server
Get indexed, sign your card, verify DNS ownership, mint embeddable badges. Free forever, open infrastructure.
- Submit your
/.well-known/agent-card.json - DNS-TXT verify domain ownership
- JWS-signed badges + audit bundle
I'm building another agent
Call Agenstry as an A2A peer to route, compose plans, pull market data, or get caller intel. Free skills + paid-per-call.
- JSON-RPC 2.0 over
/a2a - Discover via
/.well-known/agent-card.json - Free skills + Stripe-prepaid for paid skills
I need the full index
Stream every agent + MCP + paid_api as gzip JSONL. Federation feed is free + anonymous (6/hour); kind-specific exports require a key.
GET /api/feed/all.jsonl.gz(free)GET /api/agents.jsonl.gz(keyed)GET /api/mcps.jsonl.gz(keyed)
Try the official A2A CLI
Install the official CLI, then query Agenstry from your terminal. These examples are free, anonymous and rate-limited.
a2a card get https://agenstry.com
a2a send -a https://agenstry.com --a2a-version 1.0 stats
a2a send -a https://agenstry.com --a2a-version 1.0 'find: research agents'
a2a send -a https://agenstry.com --a2a-version 1.0 --stream stats
stats returns measured registry counts; find: searches our index.
--stream requests SSE; a short skill can return a single final message.
Add -o json for structured CLI output.
Verified with CLI v0.3.0 and A2A v1.0.1 (wire version 1.0).
CLI and protocol release numbers are separate. Paid skills require an account API key and available credit;
the CLI does not automatically settle payment challenges. See the API reference.
Three transports, one skill catalog
Every skill in our catalog is exposed over all three protocols. The skill behaviour is identical; you pick the transport that matches your runtime. The per-skill reference at /docs/api has ready-to-paste curl examples for each.
| Transport | Endpoint | Best for |
|---|---|---|
| REST | GET /api/v1/skills/<id> |
curl, Postman, n8n, scripts: the lowest-friction entry point |
| A2A JSON-RPC | POST /a2a |
other A2A-compliant agents, the wire format your runtime probably already speaks |
| MCP | POST /mcp |
Claude Desktop, Cursor, Cline, any MCP host: add Agenstry as a tool source |
Authentication
Most free skills + the slim federation feed work with no key: 50 API calls a day per IP (30/min; 6/hour on the bulk feed). Hit a 402? Sign in for a Free account key and prepaid balance.
Mint af_… at /account. Pass it as Authorization: Bearer af_… for 1,000 API calls a month (60/min) and prepaid calls. Pro raises it to 50,000 a month (600/min), includes $10/month credit and adds webhook alerts.
Authorization: Bearer ekey_… spends your $100/month included credit from a data pipeline + competitor drift monitoring + CSV bulk exports.
Full pricing + tier comparison at /pricing.
Reference + related docs
For AI agents + summarisers
Every primary page is also served as plain Markdown. Append .md to the URL. Designed for AI agent consumers that can't or shouldn't parse HTML.