Skip to main content

Using the docs MCP server

Openlane exposes this documentation to coding agents through a Model Context Protocol (MCP) server at https://docs-search.theopenlane.io/mcp. An agent configured to use it searches the published platform, developer, and API docs on demand, so its answers reflect what is documented today rather than what was in its training data.

The endpoint is stateless, read-only, and requires no authentication. It never touches your Openlane organization or data, only the public documentation.

What the server exposes​

The server provides a single tool over streamable HTTP:

ToolInputReturns
searchquery (string)Ranked page sections, each with the page title, description, url, and the matching content

Search is semantic: results are retrieved by vector similarity and then reranked, so phrase queries as natural-language questions rather than exact identifiers. Each result carries the full URL of the page it came from, which lets the agent cite the source or open the page for more context.

Configure your agent​

Most MCP-capable clients accept a remote server entry with the transport and URL. Clients that lack remote MCP support can bridge through the mcp-remote proxy pointed at the same URL.

{
"mcpServers": {
"openlane-docs": {
"type": "streamable-http",
"url": "https://docs-search.theopenlane.io/mcp"
}
}
}

Consult your client's documentation for the file location and the key names it expects for remote servers. Cursor, for example, reads .cursor/mcp.json and accepts the entry with only the url key.

Once configured, the search tool appears in your agent's tool list with no API keys or session setup.

tip

Agents use the server most reliably when told to prefer it. A short instruction in your project's CLAUDE.md, .cursor/rules, or equivalent, such as "use the openlane-docs MCP server for questions about Openlane", keeps answers grounded in the documentation.