Agent Papers
Get the API
HomeDeveloper access

Developer access

Use the registry without an API key. Responses are source-linked, CORS-open, and derived from the same canonical records as the site.

Choose a surface

REST API

Filter companies and inspect categories, standards, matrices, and changes.

/api/v1/companies

OpenAPI 3.1

Generate clients from stable operation IDs and response schemas.

/api/v1/openapi.json

MCP server

Query the registry with tools for companies, categories, standards, and changes.

/api/mcp

Registry JSON

Download the complete canonical registry in one request.

/registry.json

Registry SHA-256

Verify the exact bytes of a downloaded registry snapshot.

/registry.json.sha256

Registry schema

Inspect and validate the registry structure with JSON Schema Draft 2020-12.

/registry.schema.json

Registry CSV

Load one row per offering into a spreadsheet or data tool.

/registry.csv

Full registry text

Read all source-linked records as one plain-text document.

/llms-full.txt

Markdown mirrors

Fetch individual company and category pages as Markdown.

/companies/agentmail.md

RSS feed

Follow additions and re-checks with stable item identifiers.

/feed.xml

JSON Feed

Consume the same registry changes as structured JSON Feed 1.1 items.

/feed.json

Start with REST

The API is read-only. Successful and error responses use documented JSON envelopes. The OpenAPI document is the contract for generated clients, the registry schema describes the complete download's structure, and the SHA-256 file verifies the downloaded bytes.

All companies
curl https://agentpapers.dev/api/v1/companies

Money providers
curl 'https://agentpapers.dev/api/v1/companies?category=money'

Search capabilities
curl 'https://agentpapers.dev/api/v1/companies?q=wallet'

Category matrix
curl https://agentpapers.dev/api/v1/categories/money/matrix

OpenAPI document
curl https://agentpapers.dev/api/v1/openapi.json

Canonical data and freshness

Each company record includes its verification date and official sources. Unknown values remain unknown rather than being inferred. Qualification and semantic validation rules are documented in the methodology.