Skip to main content
Glama
mi7chal

repomap-mcp

by mi7chal

repomap-mcp

Local stdio Model Context Protocol (MCP) server for Aider's AST Tree-sitter and PageRank repository mapping.

Deterministic, offline, zero API keys. Generates concise code skeletons to orient LLMs within token budgets.

Setup

Add to your MCP client config (Claude Desktop, Cursor, Cline):

{
  "mcpServers": {
    "repomap": {
      "command": "uv",
      "args": ["run", "--directory", "/path/to/repomap-mcp", "repomap-mcp"]
    }
  }
}

Or run directly from GitHub without cloning:

{
  "mcpServers": {
    "repomap": {
      "command": "uv",
      "args": [
        "run",
        "--from",
        "git+https://github.com/mi7chal/repomap-mcp",
        "repomap-mcp"
      ]
    }
  }
}

Related MCP server: codebase-rag

Tools

get_repo_map

Returns a PageRank-ranked code skeleton fitting within max_map_tokens, prefixed with coverage telemetry.

  • root_path (string, default: "."): Target directory (scopes candidate files; "." maps entire repo).

  • include_tests (bool, default: false): Include unit tests and tests/ directories.

  • types_only (bool, default: false): Filter to type definitions (structs, classes, interfaces, enums).

  • focus_files (list[string]): Currently active/edited files to bias PageRank towards.

  • other_files (list[string]): Explicit file list. Defaults to scoped discovered files.

  • mentioned_idents (list[string]): Symbol names to prioritize.

  • max_map_tokens (int, default: 2048): Max token budget.

  • force_refresh (bool, default: false): Invalidate tag cache.

get_tags

Extracts Tree-sitter definitions and references for a specific file.

  • file_path (string): Target source file.

  • root_path (string, default: "."): Repository or target directory.

Cache

  • Git repos: .git/repomap-mcp/cache (natively ignored by git, zero root pollution).

  • Non-git / read-only: ~/.cache/repomap-mcp/tags-<hash>.

  • Override: REPOMAP_CACHE_DIR environment variable.

Development

uv run pytest
uv run prek run --all-files

AI Transparency

This project was developed with AI assistance.

License

Apache License 2.0. Derived from Aider by Paul Gauthier. See LICENSE and NOTICE.

Related MCP Connectors

Related MCP Servers

  • A
    license
    A
    quality
    D
    maintenance
    Enables LLMs to efficiently navigate large codebases by providing surgical access to specific code symbols via semantic search and call-graph queries.
    6
    MIT
  • A
    license
    Not graded
    quality
    D
    maintenance
    Enables LLM agents to efficiently understand and navigate a codebase by providing semantic search over symbols and a reference graph, replacing expensive grep/glob calls with structured tools like definition lookup, caller/callee queries, and change-impact analysis.
    3
    MIT
  • F
    license
    Not graded
    quality
    B
    maintenance
    Enables LLMs to analyze Python repositories structurally via AST and call graphs, providing code search, symbol lookup, change-impact analysis, workflow tracing, and explanation without embeddings.
    1
    -
  • A
    license
    Not graded
    quality
    C
    maintenance
    Enables AI agents to explore a repository by querying a pre-built structural knowledge graph of symbols and their relationships — definitions, callers, callees, blast radius, and file-level impact — instead of grepping and reading files. This cuts the tokens spent on code exploration by roughly 96–99% while running entirely on local SQLite with tree-sitter parsing.
    1
    MIT