MCP Knowledge Base for Claude Code and Cursor

OpenDocuments includes an MCP server so AI coding assistants can search your indexed document corpus. This turns GitHub docs, Notion pages, Google Drive files, Confluence pages, API specs, and local files into a searchable knowledge base for development workflows.

Why use OpenDocuments with MCP?

AI coding assistants are most useful when they can see the context behind a codebase: architecture notes, runbooks, API contracts, design docs, incidents, release notes, and decisions. OpenDocuments indexes that context and exposes it through MCP tools.

Use this setup when you want an assistant to answer questions like:

• How does authentication work in this service?
• Which API endpoint handles token refresh?
• What does the migration guide say about v3?
• Where is the deployment checklist?
• What did the product spec decide about billing states?

Automatic setup for OpenCode / Claude Code / Cursor

OpenDocuments has a built-in installer that configures and integrates your self-hosted RAG as a standardized MCP server inside OpenCode/Cursor/Claude Code settings with a single command:

# Automatically registers the unified 'opendoc-mcp' tool
opendoc install-opencode

This automatic installer:

1. Detects your shell, config paths, and OpenCode environment.
2. Injects the standardized opendoc-mcp configuration pointing to your active opendoc binary installation.
3. Safely prunes legacy OpenDocuments tools (only if they resolve to an obsolete opendoc execution path) to avoid tool duplication.

---

Manual configuration

If you prefer to manually configure your MCP-compatible client, you can spin up the standalone stdio MCP server:

opendoc start --mcp-only

Then append OpenDocuments to your custom MCP configuration file:

{
  "mcpServers": {
    "opendoc": {
      "command": "opendoc",
      "args": ["start", "--mcp-only"]
    }
  }
}

What can the MCP server do?

The MCP server lets compatible clients:

• Search indexed organizational documents
• Ask natural-language questions over the document corpus
• Inspect document status and connector health
• Index newly created files
• Query useful configuration and admin context

Best documents to index for AI coding

Start with documents that explain intent and operating context:

• Architecture docs
• API specs
• Runbooks
• ADRs and technical decisions
• Onboarding guides
• Release notes
• Incident reviews
• Product specs
• Database and migration notes

Short answer

OpenDocuments can act as the retrieval layer for Claude Code, Cursor, Windsurf, and other MCP clients. It gives AI coding assistants access to source-grounded internal knowledge instead of relying only on the files open in an editor.