MCP Docs Server: AI assistant context for GAIA development #201
Description
Summary
Create an MCP Documentation Server that gives AI coding assistants (Claude Code, Cursor, GitHub Copilot, VSCode) intelligent access to GAIA SDK documentation.
Goal: Make AI assistants expert GAIA developers out of the box.
The Problem
When developers use AI coding assistants to build GAIA agents, the AI lacks context about:
- SDK patterns and best practices
- Available tools, mixins, and base classes
- Correct usage examples
- Version-specific APIs
Current workarounds:
| Approach | Limitation |
|---|---|
| Web fetch | Slow, one page at a time, requires network |
| Static CLAUDE.md | Gets outdated, limited coverage |
| Copy-paste docs | Manual, context window bloat |
The Solution
An MCP server that exposes GAIA documentation as tools that AI assistants can call:
gaia mcp docs startKey Features
- Semantic search - Find relevant docs across all pages with natural language queries
- Code examples on demand - Get working snippets for any SDK pattern
- Version aware - Returns docs matching installed GAIA version
- Works everywhere - Claude Code, VSCode, Cursor, and any MCP-compatible client
MCP Tools
| Tool | Description |
|---|---|
search_docs |
Semantic search across all GAIA documentation |
get_page |
Retrieve full content of a specific doc page |
get_code_example |
Get working code for patterns like basic-agent, mcp-agent, rag-query |
list_components |
List available agents, mixins, tools, and SDKs |
Usage
Start the server:
gaia mcp docs startClaude Code config:
{
"mcpServers": {
"gaia-docs": {
"command": "gaia",
"args": ["mcp", "docs", "start"]
}
}
}Success Metrics
| Metric | Target |
|---|---|
| Search latency | < 100ms |
| Index build time | < 30 seconds |
| Query relevance (top-3 hit rate) | > 80% |
Resources
Vote with 👍 to show your interest and help us prioritize!