Documentation.
Without the guesswork.
From your first MCP connection to focused, version-aware retrieval. Everything below maps to Docsy’s actual open-source tools and public API.
One endpoint.
Better context.
Docsy speaks Streamable HTTP MCP. Point a compatible client to the public endpoint, then ask it to resolve a library and retrieve documentation.
Copy the MCP endpoint
Use the JSON configuration below, or switch to the direct endpoint URL.
Connect your MCP client
Add Docsy under your client’s MCP server settings and choose Streamable HTTP when prompted.
Ask a version-specific question
Resolve a library, then use query-docs to request grounded excerpts with source links.
{
"mcpServers": {
"docsy": {
"url": "https://docsy.remcp.site/mcp"
}
}
}Works with your tools.
Use the connection configuration above with any editor or agent that supports remote Streamable HTTP MCP servers. Exact settings vary by client; the Docsy CLI can generate configurations for supported editors.
resolve-library-idResolve package names to stable library IDs.query-docsRetrieve focused excerpts for a particular version.list-library-versionsInspect indexed and available library versions.search / fetchSearch the corpus and retrieve a result by its ID.Simple HTTP requests.
Prefer an ordinary REST call? The public retrieval API uses the same index as MCP. Resolve a library ID before querying its documentation.
Resolve a library
curl --get 'https://docsy.remcp.site/v2/libs/search' \
--data-urlencode 'libraryName=React' \
--data-urlencode 'query=React useEffect cleanup'Query its documentation
curl --get 'https://docsy.remcp.site/v2/context' \
--data-urlencode 'libraryId=/facebook/react' \
--data-urlencode 'query=How do I clean up useEffect?' \
--data-urlencode 'type=json'Set type=txt for agent-ready plain text. Set type=json for structured results and original source links. Use a versioned library ID such as /org/project/v1.2.3 when available.
From terminal to code.
The open-source CLI and SDK are available in the Docsy repository ↗. The CLI includes setup commands for several MCP clients and a direct documentation search workflow.
docsy setup --cursor
docsy setup --claude
docsy setup --codex
docsy library react "useEffect cleanup"
docsy docs /facebook/react "useEffect cleanup"TypeScript SDK
import { Docsy } from "@docsy/sdk";
const docsy = new Docsy({
baseUrl: "https://docsy.remcp.site",
apiKey: process.env.DOCSY_API_KEY
});
const libraries = await docsy.searchLibrary(
"How do I configure middleware?", "Express"
);
const docs = await docsy.getContext(
"JWT middleware example", libraries[0].libraryId
);See the repository’s CLI guide ↗ and SDK guide ↗ for setup and additional options.
Your context.
Your control.
Is public retrieval free?
Yes. Public MCP and REST retrieval support anonymous, read-only requests with rate limits. No paid model or API subscription is needed for the lexical retrieval mode.
How do API keys work?
Sign in with Google to create and revoke a personal key. Use Authorization: Bearer docsy_… to attribute API/MCP usage to your dashboard.
Can I self-host?
Yes. The crawler, index, MCP server, API and frontend are in the MIT-licensed repository ↗. Storage and ingestion run in your own environment.