Query the Sigistry catalog of verified Claude Code plugins and portable skills over plain HTTP and JSON. The same data the site and the MCP server use, described by an OpenAPI 3.1 spec so you can wire it into a custom GPT, a code generator, or any tool that speaks OpenAPI. No key, no signup.
https://sigistry.com/v1/openapi.jsonPrefer MCP? The same catalog is a read-only MCP server at sigistry.com/mcp.
Base URL https://sigistry.com. Every route is a GET, returns JSON, and carries Cache-Control: public, max-age=300.
/v1/pluginsList and search plugins. Query: q (keyword), category, limit, offset.
/v1/plugins/{id}One plugin by id, with its full verification object (per-check results).
/v1/verified/{id}The security-audit record for one plugin: status, checks, badge, pinned commit.
/v1/skillsList and search verified skills. Query: q, plugin, limit, offset.
/v1/skills/{name}One skill by name, including the full portable SKILL.md source.
/v1/categoriesEvery plugin category with its count, plus the total.
/v1/openapi.jsonThe OpenAPI 3.1 spec for this API.
Search for database plugins:
curl https://sigistry.com/v1/plugins?q=database{
"total": 1,
"count": 1,
"limit": 50,
"offset": 0,
"plugins": [
{
"id": "sql-safety-net",
"name": "SQL Safety Net",
"category": "database",
"installCommand": "/plugin install sql-safety-net@sigistry",
"verificationStatus": "verified"
}
]
}Two ways into the same catalog. Pick the one your tool already speaks.
Model Context Protocol over Streamable HTTP at sigistry.com/mcp. Best for agents with native MCP support (Claude Code, Claude Desktop). Tools for search, detail, categories, skills, and local verification.
Plain HTTP and JSON under /v1, described by OpenAPI 3.1. Best for custom GPTs, code generators, dashboards, and anything that ingests a spec rather than MCP.
Yes, free, and no. It is a public, read-only, authless API over data that is already public. There is no signup, no token, and no per-key rate limit. Please be reasonable with request volume; responses are cacheable for five minutes.
Same catalog data, two shapes. The MCP server (sigistry.com/mcp) speaks the Model Context Protocol over Streamable HTTP, ideal for agents like Claude Code that natively support MCP. The REST API speaks plain HTTP and JSON and is described by OpenAPI, which is what custom GPTs, code generators, and most integration tools expect. Use whichever your tool consumes.
Yes. Point your GPT action or tool at the OpenAPI document at https://sigistry.com/v1/openapi.json and it can call the endpoints directly. The spec is OpenAPI 3.1.
The catalog and verification results are read from the marketplace repository and cached about five minutes, so a newly merged or re-verified plugin appears within roughly that window.
verified means the plugin passed the eight-check security methodology; listed means it is in the registry but not audited; stale means it was verified at a pinned commit the repo has since moved past; failed means an audit check failed; unknown means there is no record. Detail endpoints return the full per-check breakdown.