Cortex MCP Server has 5 tools with INCONSISTENT quality. All tools have descriptions and input schemas are present with type definitions, but several lack critical parameter documentation and output schema specification. Tool naming follows verb_noun conventions (cortex_search, cortex_get_*), which is good. However, parameter descriptions are sparse or absent, and there is NO documented output schema for any tool. The server appears functional but falls short of production-grade agent tool standards. Schema coverage is incomplete, parameter descriptions are minimal, limiting LLM's ability to reason about valid inputs. No error handling guidance visible in tool definitions.

• Document the output schema for each tool. Specify which fields are returned, their types, and which ones are required. Example for cortex_search: { results: [{ id, title, content, branch, score }], total_count, next_cursor }
• Expand parameter descriptions to include expected format, valid ranges, and when to use them. Example: 'limit (integer, 1-100, default 10): Maximum number of results to return. Use lower values for quick filtering; higher for comprehensive results.'
• Add constraints to numeric parameters: limit should be 1-100 (default 10); similar bounds for all count/limit params across tools
• Clarify the 'branch' parameter: is it a filter or a required selection? Document what 'current branch' resolves to and how to override it.
• Add error guidance to tool descriptions. Example: 'If no results found, try broadening the query or check if the branch exists. Call cortex_get_status() first to verify project initialization.'
• Distinguish cortex_search (general hybrid search) from cortex_search_decisions (decisions only) in descriptions. Example: 'Use cortex_search for general memory queries. Use cortex_search_decisions specifically to find immortal decisions or rejections.'
• Document what cortex_get_status returns. Currently unclear what 'status and capabilities' includes. List specific fields: { initialized, tier, has_embeddings, branch_count, decision_count, etc. }
• Consider adding a dependency note: 'Requires cortex to be initialized in the current project. Call cortex_get_status() first to verify.' This prevents errors on uninitialized projects.