MCP Tool Reference

When running the CodeGraphContext (CGC) MCP server, it registers a suite of 25 JSON-RPC tools that AI assistants can use to analyze and query the code graph.

Code Ingestion & System Control

`add_code_to_graph`

Indexes a local directory or file into the active database context. - Parameters: - path (string, required): Absolute filesystem path. - is_dependency (boolean, optional): Marks the code as an external library.

`add_package_to_graph`

Discovers, downloads (if needed), and indexes a third-party package. - Parameters: - package_name (string, required): Name of the package (e.g., requests, express). - language (string, required): Language syntax parser (python, javascript, typescript, java, c, go, ruby, php, cpp). - is_dependency (boolean, optional): Marks the package as an external library (default: true).

`watch_directory`

Launches a directory watcher for incremental updates. - Parameters: - path (string, required): Directory path to watch.

`unwatch_directory`

Stops file monitoring on a folder path. - Parameters: - path (string, required): Directory path.

`list_watched_paths`

Lists all directories currently monitored by watchers.

`delete_repository`

Removes a repository's code structures from the graph. - Parameters: - repo_path (string, required): Repository path.

`list_indexed_repositories`

Returns a list of all repositories stored in the active database.

`get_repository_stats`

Retrieves ingestion metrics (counts of files, functions, classes, modules). - Parameters: - repo_path (string, optional): Restricts stats to a specific repository.

Background Job Controller

Some operations (like indexing large codebases) execute as background tasks. Use these tools to monitor task states:

`list_jobs`

Lists all background jobs and their execution states.

`check_job_status`

Queries progress and logs for a specific job. - Parameters: - job_id (string, required): Target job ID.

Code Search & Relationship Analysis

`find_code`

Searches symbol definitions, file names, or source code for keyword matches. - Parameters: - query (string, required): Target keyword or pattern. - fuzzy_search (boolean, optional): Enables fuzzy matching. - edit_distance (number, optional): Levenshtein distance limit (0-2). - repo_path (string, optional): Restricts search scope.

`analyze_code_relationships`

The primary tool for traversing structural relationships in the graph. - Parameters: - query_type (string, required): The traversal type. Must be one of: - find_callers: Find immediate caller functions. - find_callees: Find immediate functions called by target. - find_all_callers: Deep search up the invocation chain. - find_all_callees: Deep search down the execution path. - find_importers: Find files importing the target symbol/module. - who_modifies: Trace variables or structures written to. - class_hierarchy: Resolves superclass and subclass trees. - overrides: Finds functions overriding parent methods. - dead_code: Scans context for unused subroutines. - call_chain: Traces invocation chains between source and destination. - module_deps: Identifies dependencies between modules. - variable_scope: Tracks variable bindings. - find_complexity: Returns cyclomatic complexity score. - find_functions_by_argument: Searches for functions declaring target parameter. - find_functions_by_decorator: Searches for functions decorated with target. - target (string, required): The identifier name to analyze. - context (string, optional): Specific file path to resolve target namespace conflicts. - repo_path (string, optional): Restricts search scope.

`calculate_cyclomatic_complexity`

Computes the complexity score of a function. - Parameters: - function_name (string, required): Function identifier. - path (string, optional): File path containing definition. - repo_path (string, optional): Restricts search scope.

`find_most_complex_functions`

Returns methods with the highest cyclomatic complexity scores. - Parameters: - limit (integer, optional): Maximum rows to return (default: 10). - repo_path (string, optional): Restricts search scope.

`find_dead_code`

Scans for unreferenced code declarations. - Parameters: - exclude_decorated_with (array of strings, optional): Excludes functions carrying specified decorator annotations (e.g., @app.route). - repo_path (string, optional): Restricts search scope.

Workspace Context Management

`discover_codegraph_contexts`

Scans subdirectories for existing .codegraphcontext/ directories. - Parameters: - path (string, optional): Scan root directory. - max_depth (integer, optional): Folder depth limit (default: 1).

`switch_context`

Reconnects the MCP session to a different graph database. - Parameters: - context_path (string, required): Path to the target repository root containing .codegraphcontext/. - save (boolean, optional): Persists configuration mapping (default: true).

Advanced Querying & Reporting

`execute_cypher_query`

Executes raw Cypher queries directly against the graph database. - Parameters: - cypher_query (string, required): Cypher statement.

`visualize_graph_query`

Generates a Neo4j visualization link. - Parameters: - cypher_query (string, required): Cypher statement to render.

`generate_report`

Compiles a markdown quality report (CGC_REPORT.md). - Parameters: - output_path (string, optional): Output report path. - include_java (boolean, optional): Appends Spring endpoints and bean tables. - god_node_limit (integer, optional): Limit for high fan-in symbol rows. - complexity_limit (integer, optional): Limit for complex method rows. - cross_module_limit (integer, optional): Limit for module coupling rows.

Portability & Registries

`load_bundle`

Imports a portable .cgc file, downloading it from the registry if needed. - Parameters: - bundle_name (string, required): Package name (e.g., requests) or file name. - clear_existing (boolean, optional): Purges active context before loading.

`search_registry_bundles`

Searches the public CGC bundle server database. - Parameters: - query (string, optional): Name or description keywords. - unique_only (boolean, optional): Returns only the latest version of packages.

Frameworks & Datasource Extensions

`find_java_spring_endpoints`

Searches Spring controller REST mappings. - Parameters: - http_method (string, optional): Method filter (GET, POST, etc.). - path_pattern (string, optional): URL path substring. - repo_path (string, optional): Restricts search scope.

`find_java_spring_beans`

Returns registered Spring stereotype beans. - Parameters: - stereotype (string, optional): Stereotype filter (SERVICE, REPOSITORY, etc.). - repo_path (string, optional): Restricts search scope.

`find_datasource_nodes`

Returns ingested MySQL, Cassandra, or Redis schema nodes. - Parameters: - kind (string, optional): Datasource kind (mysql, cassandra, redis). - name (string, optional): Substring filter for names. - include_columns (boolean, optional): Appends table columns/key patterns (default: false).