MCP Tools and Resources Reference

This document provides complete reference specifications for all 14 Model Context Protocol (MCP) tools, dynamic resources, and guided prompts exposed by ContextCortex.

---

MCP Tools Matrix

Tool Name	Minimum Role	Primary Parameters	Description
search_code	viewer	query, repo, language, limit	Performs hybrid dense and sparse search over indexed source code chunks.
search_docs	viewer	query, repo, category, tag, limit	Searches markdown documents, system architecture notes, and runbooks.
find_symbol	viewer	name, repo, exact, limit	Looks up AST symbols (functions, classes, interfaces) with exact or prefix matching.
get_file_outline	viewer	filepath, repo	Returns the structural outline of a file without full token payload costs.
list_repositories	viewer	None	Lists all registered Git repositories and local monitored paths with sync status.
sync_repository	editor	repo	Initiates background shallow clone synchronization for a repository.
index_status	viewer	None	Returns vector collection totals, embedding model details, and GitHub rate limits.
get_architecture	viewer	repo	Synthesizes repository entry points, language distributions, and structural layouts.
manage_adr	editor	action, repo, title, decision, status	Lists, creates, or updates Architectural Decision Records (MADR format).
get_code_routes	viewer	repo, framework, http_method	Returns declared HTTP route definitions and endpoint handlers across backend frameworks.
trace_call_path	viewer	target, repo, direction, depth	Traces AST symbol calls, imports, and cross-repo connections via breadth-first search.
manage_local_file	editor / viewer	action, file_path, content, repo	Manages files in local storage: upload, replace, read, or delete with real-time indexing.
what_is_ingested	viewer	source_type, repo_name, path_prefix, detail_level	Inspects all ingested sources with multidimensional filtering and file tree hierarchies.

---

Detailed Tool Specifications

1. search_code

Performs hybrid semantic and BM25 search over indexed code blocks.

• Parameters:
  • query (string, required): The natural language or code snippet search query.
  • repo (string, optional): Restrict results to a specific repository alias.
  • language (string, optional): Filter by programming language (e.g. python, typescript).
  • limit (integer, default: 5): Maximum number of ranked results to return.
• Returns: Array of code chunks with file paths, line ranges, RRF scores, and clickable Git permalinks.

2. find_symbol

Deterministic symbol lookup from the AST index.

• Parameters:
  • name (string, required): Symbol name to find (e.g. FastEmbedEngine or search_code).
  • repo (string, optional): Repository alias filter.
  • exact (boolean, default: true): Perform exact matching when true, prefix search when false.
  • limit (integer, default: 10): Maximum matches to return.
• Returns: Symbol declarations with signatures, starting and ending lines, and docstrings.

3. manage_local_file

Uploads, reads, updates, or deletes files in the managed local storage directory.

• Parameters:
  • action (string, required): Operation to execute (upload, replace, read, delete).
  • file_path (string, required): Relative file path within storage. Path traversal tokens (..) are rejected.
  • content (string, optional): Text file content for upload or replace actions.
  • repo (string, optional): Logical repository namespace.
• Returns: File status, file size in bytes, and indexing confirmation.

4. what_is_ingested

Returns an inventory of all indexed sources in the system.

• Parameters:
  • source_type (string, optional): Filter by all, git, monitored_path, or local_storage.
  • repo_name (string, optional): Filter by specific repository.
  • path_prefix (string, optional): Path prefix filter.
  • file_extension (string, optional): File extension filter (e.g. .py, .ts).
  • detail_level (string, default: summary): Detail level (summary or detailed).
• Returns: Aggregated file counts, source statuses, or hierarchical directory trees.

---

Dynamic Catalog Resources

ContextCortex exposes dynamic resources for AI agents:

Resource URI	MIME Type	Description
knowledge://catalog/summary	text/markdown	Real-time markdown report summarizing all indexed repositories, document distributions, and AST symbol totals.

---

Guided Prompts

ContextCortex exposes guided workflows via MCP prompts:

1. search_infrastructure_docs

• Arguments: topic (string)
• Description: Guides an AI assistant to retrieve container configurations, port mappings, and reverse proxy routes.

2. find_implementation_symbol

• Arguments: symbol (string), repo (string, optional)
• Description: Guides an AI assistant to locate symbol definitions, parameter signatures, and implementations.