Multi-repo semantic code search MCP server with hybrid vector+BM25 retrieval, fully offline.
codesearch is worth checking the docs before setup with strong trust signals. Check agent compatibility and use-case fit before adding it to your workflow.
gh repo view flupkede/codesearch --webOpen the official README and confirm the supported install method.
Add the server entry to your MCP client config.
Restart your agent and verify that the server tools appear.
Repository setup guidance
Strong trust signals; still review the README and permissions before production use.
Last commit was about 96 days ago.
38 GitHub stars indicate community interest.
0 open issues signal maintenance load.
Apache-2.0 license detected.
0 security/trust notes recorded.
Setup difficulty is 3/5.
Connect any AI model to 1200+ integrations (MCP, CLI, API)
An AI Gateway, registry, and proxy that sits in front of any MCP, A2A, or REST/gRPC APIs, exposing a unified endpoint with centralized discovery, guardrails and management. Optimizes Agent & Tool calling, and supports plugins.
Official Microsoft Learn MCP server and CLI tool providing real-time, trusted Microsoft docs and code samples to AI agents.
Open-source credential gateway with a built-in vault. give your AI agents access to services without exposing keys.
codesearch is a tool that helps AI agents search and understand code across multiple repositories at once. It uses both semantic meaning and exact text matching to find relevant code, and it runs entirely on your machine without needing internet or a GPU. You can index your code once and then ask questions like 'find the function that handles authentication' or 'show me all usages of this class'.
codesearch is a Rust-based MCP (Model Context Protocol) server that provides multi-repository semantic code search for AI agents. It combines vector embeddings (ANN) with BM25 full-text search using Reciprocal Rank Fusion (RRF) for hybrid retrieval. Key features include: multi-repo serve mode with cross-repo RRF ranking, symbol navigation (definitions, usages, imports, dependents), AST-aware chunking via tree-sitter for 15 languages, token-efficient responses (compact by default, full code fetched on demand), and a lightweight footprint (hundreds of MB on disk, CPU-only, no Docker). It offers 6 MCP tools: search (semantic or literal), find (symbol navigation), explore (outline or similar), get_chunk (retrieve full code), find_impact (C# impact analysis via SCIP), and status. Designed for enterprise/offline use, it requires no runtime model downloads and works behind proxies. Indexing is zero-config for single repos: just run 'codesearch index && codesearch mcp'.
38
Stars
9
Forks
0
Issues
Apache-2.0
License
Multi-repo semantic code search MCP server with hybrid vector+BM25 retrieval, fully offline.
AI agent developers building coding assistants, Software engineers working with large multi-repo codebases, DevOps engineers needing offline code search in air-gapped environments
The setup section provides repository-level starting guidance, not a guarantee of an independently verified installation. Check the official README and release notes.
Review repository permissions, executed commands, data sent to external services, and dependencies before use.
Similar or complementary options to evaluate include metorial, mcp-context-forge, Microsoft Learn MCP Server.
Aug 20, 2026