Codegraphcontext

What shashankss1205/codegraphcontext MCP server does

The shashankss1205/codegraphcontext MCP server converts a local code repository into a queryable knowledge graph. Instead of representing a project only as searchable text, it records structural elements such as files, symbols, imports, function calls, callers, callees, class hierarchies, and relationships between modules.

AI assistants can access this indexed repository through MCP and use the graph to investigate code structure. Developers can use the accompanying CLI for repository analysis without an AI client. This makes the project suitable for tracing cross-file call chains, examining inheritance, checking module boundaries, identifying potential dead code, and investigating dependencies.

The project also supports pre-indexed .cgc bundles for repositories that have already been analyzed. A bundle can be loaded without repeating the indexing process.

How it works

Source files are processed by Tree-sitter-based indexers, with SCIP indexing available as an optional path. The indexers extract code entities and their relationships, then store that information in a graph database. The resulting graph is available to both the CLI and the MCP server.

The shashankss1205/codegraphcontext MCP server is one operating mode of the project; the other is a standalone command-line toolkit. A live-watching mode can monitor directories and update the graph when files change. This is useful when the repository is being edited while an assistant or developer continues querying it.

The project supports 23 programming languages. Its documented database choices include FalkorDB Lite as the default, KuzuDB, LadybugDB, FalkorDB Remote, Nornic DB, and Neo4j. The README states that CGC can work locally with embedded backends or use external graph databases when needed.

Setup and configuration

Installation is provided through the Python package published as codegraphcontext. The project includes an interactive command-line setup flow intended to help configure a local installation. After setup, a repository must be indexed before relationship queries can provide repository-specific results, unless a suitable .cgc bundle is loaded instead.

The default database arrangement is FalkorDB Lite. Alternative backends can be selected when the deployment requires a remote or different graph database, including Neo4j through Docker or native installation. The supplied material does not specify required environment variable names, a complete MCP client configuration block, or an exact server-launch command.

Tools and capabilities

The shashankss1205/codegraphcontext MCP server supports repository-context workflows involving:

• Querying callers, callees, and call chains.
• Inspecting class hierarchies and imports.
• Exploring relationships across files and packages.
• Providing graph-based repository context to AI assistants.
• Keeping an indexed directory current through live file watching.
• Loading pre-indexed .cgc repository bundles.

The CLI exposes the same indexed graph for direct code analysis, so the project can be used without MCP when an agent integration is not required.

Limitations and notes

An indexing step is part of the normal workflow, which adds preparation time compared with plain file search. Results depend on the repository having been indexed and on the supported language and indexer handling its source code. The README describes broad language support but does not provide language-specific coverage details in the supplied material.

The shashankss1205/codegraphcontext MCP server is intended for local code context and graph analysis; the provided information does not describe hosted indexing, access control, authentication, or synchronization across users. Database selection may also add operational requirements when using external backends rather than the embedded default.