Skip to content

MCP_TOOLS.md

Latest commit

 

History

History
137 lines (103 loc) · 3.9 KB

MCP_TOOLS.md

File metadata and controls

137 lines (103 loc) · 3.9 KB

MCP Tools Reference

This document describes the Model Context Protocol (MCP) tools available in CodeGraphContext. These tools are exposed to AI assistants to enable them to interact with your codebase.

📋 Tool Categories

  1. Indexing & Management
  2. Code Search
  3. Analysis & Quality
  4. Bundle Management
  5. Monitoring
  6. Advanced Querying

Indexing & Management

add_code_to_graph

Performs a one-time scan of a local folder to add its code to the graph. Ideal for libraries or dependencies.

  • Args: path (string), is_dependency (boolean)
  • Returns: Job ID

add_package_to_graph

Add an external package to the graph by discovering its location.

  • Args: package_name (string), language (string), is_dependency (boolean)
  • Returns: Job ID
  • Supported Languages: python, javascript, typescript, java, c, go, ruby, php, cpp

list_indexed_repositories

List all repositories currently indexed in the graph.

  • Args: None
  • Returns: List of repositories with paths and stats

delete_repository

Delete a repository from the graph.

  • Args: repo_path (string)
  • Returns: Success message

check_job_status

Check the status of a background job (indexing, scanning).

  • Args: job_id (string)
  • Returns: Job status (running, completed, failed) and progress

list_jobs

List all background jobs.

  • Args: None
  • Returns: List of all jobs

Code Search

find_code

Find code snippets related to a keyword.

  • Args: query (string), fuzzy_search (boolean), edit_distance (number)
  • Returns: Matches with file path, line number, and content

Analysis & Quality

analyze_code_relationships

Analyze relationships between code elements.

  • Args:
    • query_type (enum): find_callers, find_callees, class_hierarchy, overrides, etc.
    • target (string): Function/class name
    • context (string): Optional file path
  • Returns: List of related elements

find_dead_code

Find potentially unused functions across the codebase.

  • Args: exclude_decorated_with (list of strings)
  • Returns: List of unused functions

calculate_cyclomatic_complexity

Calculate complexity for a specific function.

  • Args: function_name (string), path (string)
  • Returns: Complexity score

find_most_complex_functions

Find the most complex functions in the codebase.

  • Args: limit (integer)
  • Returns: List of functions sorted by complexity

get_repository_stats

Get statistics about indexed repositories.

  • Args: repo_path (string, optional)
  • Returns: Counts of files, functions, classes, modules

Bundle Management

load_bundle

Load a pre-indexed .cgc bundle. Can download from registry if not found locally.

  • Args: bundle_name (string), clear_existing (boolean)
  • Returns: Success message and stats

search_registry_bundles

Search for bundles in the registry.

  • Args:
    • query (string, optional): Search term
    • unique_only (boolean): Show only latest version per package
  • Returns: List of available bundles

Monitoring

watch_directory

Continuously monitor a directory for changes and keep the graph updated.

  • Args: path (string)
  • Returns: Job ID for initial scan

list_watched_paths

List directories being watched.

  • Args: None
  • Returns: List of paths

unwatch_directory

Stop watching a directory.

  • Args: path (string)
  • Returns: Success message

Advanced Querying

execute_cypher_query

Run a direct read-only Cypher query against the graph.

  • Args: cypher_query (string)
  • Returns: Raw query results

visualize_graph_query

Generate a URL to visualize query results in Neo4j Browser.

  • Args: cypher_query (string)
  • Returns: URL