ExpertPack MCP
Serve ExpertPack knowledge packs with hybrid search and traceable sources
- Platforms
- CLI, self-hosted
- Runs
- Locally
- Version
- 0.7.0
- Listing updated
- today
Similar MCP servers
About ExpertPack MCP
ExpertPack MCP is an MCP server for knowledge and memory. It runs locally on your machine and is started by your AI client.
ep-mcp turns an ExpertPack, a Markdown-based pack of expert knowledge, into a queryable service for MCP clients. It targets knowledge that frontier models tend to lack, such as support lore and product gotchas, organised as one concept per file.
Retrieval blends BM25 with vector search, adds metadata boosting, MMR re-ranking and optional graph traversal, and expands declared requirements automatically. If the embedding provider fails it falls back to lexical search. Results carry IDs, content hashes, verification dates and source files, and an optional reconstruct mode returns line ranges and original text so answers can be proven.
It installs from PyPI and uses a Gemini or Azure OpenAI key for embeddings. The README reports a 97.6 percent hit rate on a 22-question benchmark.
Install
Run in a terminal
claude mcp add ep-mcp --env GEMINI_API_KEY='<GEMINI_API_KEY>' --env AZURE_OPENAI_API_KEY='<AZURE_OPENAI_API_KEY>' -- uvx ep-mcpAdd to claude_desktop_config.json
{
"mcpServers": {
"ep-mcp": {
"command": "uvx",
"args": [
"ep-mcp"
],
"env": {
"GEMINI_API_KEY": "<GEMINI_API_KEY>",
"AZURE_OPENAI_API_KEY": "<AZURE_OPENAI_API_KEY>"
}
}
}
}Add to ~/.cursor/mcp.json
{
"mcpServers": {
"ep-mcp": {
"command": "uvx",
"args": [
"ep-mcp"
],
"env": {
"GEMINI_API_KEY": "<GEMINI_API_KEY>",
"AZURE_OPENAI_API_KEY": "<AZURE_OPENAI_API_KEY>"
}
}
}
}Add to .vscode/mcp.json
{
"servers": {
"ep-mcp": {
"type": "stdio",
"command": "uvx",
"args": [
"ep-mcp"
],
"env": {
"GEMINI_API_KEY": "<GEMINI_API_KEY>",
"AZURE_OPENAI_API_KEY": "<AZURE_OPENAI_API_KEY>"
}
}
}
}Add to ~/.codeium/windsurf/mcp_config.json
{
"mcpServers": {
"ep-mcp": {
"command": "uvx",
"args": [
"ep-mcp"
],
"env": {
"GEMINI_API_KEY": "<GEMINI_API_KEY>",
"AZURE_OPENAI_API_KEY": "<AZURE_OPENAI_API_KEY>"
}
}
}
}Add to your client's MCP settings (most use this shape)
{
"mcpServers": {
"ep-mcp": {
"command": "uvx",
"args": [
"ep-mcp"
],
"env": {
"GEMINI_API_KEY": "<GEMINI_API_KEY>",
"AZURE_OPENAI_API_KEY": "<AZURE_OPENAI_API_KEY>"
}
}
}
}Using several servers? Build one config file with the MCP config generator.
Configuration
GEMINI_API_KEYGemini API key for embeddings (or set AZURE_OPENAI_API_KEY when using Azure OpenAI)AZURE_OPENAI_API_KEYAzure OpenAI key, if the config uses the Azure embedding provider
Tools this server provides (1)
Taken from the project documentation; the running server may differ.
- ep_read
- Reads a file or section from a pack, optionally with neighbouring sections.
What it does
- Hybrid search
- Combines BM25 and vector search.
- Provenance
- Results include hashes and verification dates.
- Reconstruct mode
- Returns source spans and line ranges.
- Graph expansion
- Optional multi-hop traversal.
- Freshness flags
- Marks stale or expired content.
- Lexical fallback
- Keeps working if embeddings fail.
Good for
- Give an agent product expertise
- Ground answers with citations
- Serve a person or process knowledge pack
- Measure retrieval quality
ExpertPack MCP FAQ
What does it need?
A Gemini or Azure OpenAI key for embeddings.
What if embeddings are down?
It falls back to lexical search.
Which clients work?
Claude Desktop, Cursor, Windsurf, Claude.ai and custom hosts.
What are pack types?
Person, product and process packs.
How do I add ExpertPack MCP to Claude Code?
Run this command in your terminal: claude mcp add ep-mcp --env GEMINI_API_KEY='<GEMINI_API_KEY>' --env AZURE_OPENAI_API_KEY='<AZURE_OPENAI_API_KEY>' -- uvx ep-mcp
What are the best alternatives to ExpertPack MCP?
The closest alternatives to ExpertPack MCP are Codebase Memory, XERJ, headroom, omniget and next-ai-draw-io.
Where this comes from
- Description: written from the project's README and the project's website.
- Install steps: from the official MCP Registry entry.



