Skip to main content

orgraph-mcp

Codebase knowledge graph for coding agents. Gives Claude Code, Cursor, Codex, and other MCP-compatible agents a persistent graph of any repo — call chains, topology clusters, dependency trees — on top of hybrid code search.

Quickstart

Install uv, then:

uv tool install orgraph-mcp
orgraph install

orgraph install detects installed coding agents and wires up the MCP server automatically. Open any repo in your agent and orgraph starts working — it indexes on first run, no manual setup needed.

To undo: orgraph uninstall

What agents can do

Once configured, your agent has 6 tools:

Tool What it does
search(query) Hybrid BM25+semantic search — find code by description
trace(symbol, direction, depth) Follow call chains forward (callees) or backward (callers)
get_context(file_or_symbol) Topology cluster, community, call depth, indegree — where does this fit?
find_entry_points(kind) HTTP handlers and entry surfaces; kind = "all" | "http" | "topology"
get_dependencies(file, direction, depth) Import + call dependency tree, forward or reverse
reindex(force) Re-index changed/deleted files without restarting the server

The agent picks the right tool automatically based on what you ask.

Manual usage

# Index a repo manually (optional — serve auto-indexes)
orgraph index /path/to/repo

# Check what was indexed
orgraph status /path/to/repo

# Search from the CLI
orgraph search "authentication middleware" /path/to/repo

# Start the MCP server
orgraph serve /path/to/repo

Manual MCP config

If you prefer to configure manually instead of using orgraph install:

Claude Code (~/.claude.json):

{
  "mcpServers": {
    "orgraph": {
      "command": "uvx",
      "args": ["--from", "orgraph-mcp", "orgraph", "serve", "."],
      "type": "stdio"
    }
  }
}

Cursor (~/.cursor/mcp.json):

{
  "mcpServers": {
    "orgraph": {
      "command": "uvx",
      "args": ["--python", "3.13", "--from", "orgraph-mcp", "orgraph", "serve", "."]
    }
  }
}

The server uses . as the repo path — it starts relative to wherever your agent opens the project.

How it works

  • Extraction — tree-sitter AST extractor (SCIP compiler-accurate extraction when available)
  • Graph — Kuzu embedded graph DB with Function/Class/File nodes + CALLS/IMPORTS/INHERITS edges
  • Search — semble hybrid BM25 + Model2Vec embeddings
  • Topology — BFS entry-point clustering groups files into domain clusters
  • Communities — Leiden community detection for finer-grained groupings
  • Incremental — md5 manifest tracks changes; reindex only re-extracts what changed

Eval

Measure retrieval quality against a ground truth file:

orgraph eval /path/to/repo --ground-truth queries.json --output report.json

Ground truth format:

[
  {
    "id": "auth-flow",
    "query": "how is authentication handled",
    "relevant_files": ["auth.py", "middleware.py"],
    "relevant_symbols": ["authenticate", "require_auth"],
    "query_type": "semantic"
  }
]

Reports NDCG@10, MRR, and Precision@3.

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

orgraph_mcp-0.1.28.tar.gz (295.1 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

orgraph_mcp-0.1.28-py3-none-any.whl (203.1 kB view details)

Uploaded Python 3

File details

Details for the file orgraph_mcp-0.1.28.tar.gz.

File metadata

  • Download URL: orgraph_mcp-0.1.28.tar.gz
  • Upload date:
  • Size: 295.1 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for orgraph_mcp-0.1.28.tar.gz
Algorithm Hash digest
SHA256 a335ddb0c613c68404bc67f3212d04d5d9c15cbecf7e9653e42c7cf65dc540c8
MD5 58c692a1a64be8146e11d8d7a2264849
BLAKE2b-256 bda4f3b5d0746de436c4d1535bc7df47fb5d71ad73db92c302931c20e160dc77

See more details on using hashes here.

Provenance

The following attestation bundles were made for orgraph_mcp-0.1.28.tar.gz:

Publisher: release.yml on tss-pranavkumar/orgraph

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file orgraph_mcp-0.1.28-py3-none-any.whl.

File metadata

  • Download URL: orgraph_mcp-0.1.28-py3-none-any.whl
  • Upload date:
  • Size: 203.1 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for orgraph_mcp-0.1.28-py3-none-any.whl
Algorithm Hash digest
SHA256 ded1828e2da50acb5f85b54efbda61b114e60b7906a65060842e35446cdc0ad3
MD5 5dc5f4bd44498e89106b20f8692170fd
BLAKE2b-256 6d620ab17f25fef623dd4fd4d30d93ec5260714e6138d3e5d1abead2bcc1eb71

See more details on using hashes here.

Provenance

The following attestation bundles were made for orgraph_mcp-0.1.28-py3-none-any.whl:

Publisher: release.yml on tss-pranavkumar/orgraph

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page