Hosting

Local stdio process; optional external graph database

Transport

stdio

Endpoint

Local process; no hosted endpoint documented

Authentication

No authentication is required for the default embedded database. If configuring an external Neo4j database, credentials are stored by the setup flow in ~/.codegraphcontext/.env; do not put them in the MCP record or chat.

Requirements

• Python 3.10 or later and pip, or the documented Docker image.
• The default setup uses an embedded graph database; optional backends have their own platform requirements.
• Git is optional for cloning/indexing additional repositories.

Setup steps

1. Install the maintainer-published package with pip install codegraphcontext.
2. Run codegraphcontext --help to confirm the CLI installation, then run codegraphcontext mcp setup and select the supported MCP client. The wizard stores database credentials in ~/.codegraphcontext/.env and writes client configuration.
3. Start the server with codegraphcontext mcp start. If configuring manually, use command codegraphcontext and args [mcp, start].
4. Index a project before querying its code graph. The repository documents codegraphcontext index . for indexing the current directory.
5. Reconnect the MCP client and inspect tools/list. The repository documents that indexed paths are sandboxed to the process working directory and CGC_ALLOWED_ROOTS unless the sandbox is explicitly disabled.

The repository MCP_TOOLS.md enumerates 25 tools for version 0.4.16; the current README identifies version 0.5.1 and another getting-started page reports 21 tools. The 25 names below are a documented inventory, not an observed tools/list result; see source_conflicts.

• discover_codegraph_contexts

  Find indexed CodeGraphContext databases below a directory.

  Parameters: path, max_depth
  01
• switch_context

  Select a CodeGraphContext database for the session.

  Parameters: context_path, save
  02
• add_code_to_graph

  Index a local directory into the graph.

  Parameters: path, is_dependency
  03
• add_package_to_graph

  Index an installed dependency package.

  Parameters: package_name, language, is_dependency
  04
• list_indexed_repositories

  List repositories currently indexed in the graph.

  05
• delete_repository

  Remove an indexed repository from the graph.

  Parameters: repo_path
  06
• get_repository_stats

  Get file, function, class, and module counts.

  Parameters: repo_path
  07
• find_code

  Search indexed symbols and content.

  Parameters: query, fuzzy_search, edit_distance, repo_path
  08
• analyze_code_relationships

  Query callers, callees, imports, hierarchy, and related code.

  Parameters: query_type, target, context, repo_path
  09
• find_dead_code

  Find candidate unused functions.

  Parameters: exclude_decorated_with, repo_path
  10
• calculate_cyclomatic_complexity

  Calculate complexity for a function.

  Parameters: function_name, path, repo_path
  11
• find_most_complex_functions

  Rank functions by cyclomatic complexity.

  Parameters: limit, repo_path
  12
• load_bundle

  Load a CodeGraphContext bundle.

  Parameters: bundle_name, clear_existing
  13
• search_registry_bundles

  Search the public bundle registry.

  Parameters: query, unique_only
  14
• watch_directory

  Index and monitor a directory for changes.

  Parameters: path
  15
• list_watched_paths

  List active directory watch roots.

  16
• unwatch_directory

  Stop monitoring a directory.

  Parameters: path
  17
• list_jobs

  List background indexing and scan jobs.

  18
• check_job_status

  Check the status of an indexing job.

  Parameters: job_id
  19
• execute_cypher_query

  Run read-only Cypher against the active graph backend.

  Parameters: cypher_query
  20
• visualize_graph_query

  Build a graph-browser URL for a query.

  Parameters: cypher_query
  21
• generate_report

  Generate a Markdown code graph audit report.

  Parameters: output_path, include_java
  22
• find_java_spring_endpoints

  List indexed Spring HTTP endpoints.

  Parameters: repo_path
  23
• find_java_spring_beans

  List indexed Spring bean stereotypes.

  Parameters: stereotype, repo_path
  24
• find_datasource_nodes

  Query indexed datasource architecture nodes.

  Parameters: kind, name, include_columns
  25

This server’s available tools can vary by account, permissions, or configuration. The list shown is not its full runtime catalog.

• The server only has useful repository context after a codebase is indexed.
• Indexing, watching, bundle loading, context switching, and repository deletion can change local graph/index state; only read-only discovery/search tools are suitable for harmless checks.
• Path access is sandboxed by default to the process working directory and CGC_ALLOWED_ROOTS.
• The current README and tool documentation disagree on version/tool inventory; do not label a count exact until tools/list is observed.
• This is a community project and is not an official GitHub, IDE-vendor, or language-vendor integration.
• No MCP initialize, tools/list, or representative tool call was performed for this catalog entry.

• The repository’s current README states version 0.5.1, while docs/MCP_TOOLS.md describes version 0.4.16 and 25 tools; a getting-started MCP page also reports 21 tools. The documentation inventory is therefore partial/version-sensitive.

• maintainer documentation

  Project source documents graph indexing, code search, analysis, and repository management tools.

Documentation checked 2026-10-28T00:00:00Z. Connection has not been runtime tested.