User-owned shared memory for AI agents. CLI + stdio MCP server + web UI + ingestion for a knowledge base on your own Supabase project (or fully self-hosted with Cerefox Local).
Critical gaps across all dimensions. Tool definitions are inferred from CLI command registration (deploy-server.ts) rather than explicitly visible in MCP server code. No input schemas, parameter descriptions, or output schemas are visible in the provided source. Descriptions are present but generic (avg ~80 chars, below 194-char baseline). No evidence of error handling, validation, or recovery guidance. Tools lack parameter type definitions entirely. This server appears to be a wrapper around a CLI with minimal MCP-specific implementation.
Retrieve the audit log for document operations
Retrieve a specific document by ID
Ingest documents into the knowledge base with automatic chunking and embedding
List all projects in the knowledge base
List all versions of a document
Retrieve and manage document metadata
No input schemas visible for any tool. Cannot verify parameter types, constraints, or required fields.
Tool definitions appear to be inferred from CLI command registration rather than explicitly defined in MCP server code. No explicit tool registration with MCP SDK visible.
Descriptions are generic and lack actionable context. 'Retrieve and manage document metadata' (cerefox-metadata) does not explain WHEN to use it vs cerefox-metadata-search, what fields are returned, or what 'manage' entails. Baseline: 194 chars; observed: ~35-80 chars.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 44 | 2026-07-28+ | v2 |
Search documents by metadata fields
Search documents in the knowledge base using semantic similarity and metadata filters
No parameter descriptions visible. LLMs cannot infer what cerefox-search expects (query? filters? limit?). No constraints, enums, or format guidance. Baseline: 100% of A+ tools have param descriptions.
No output schemas documented. LLMs cannot plan downstream tool calls or extract required fields (e.g., does cerefox-search return document_id, project_id, metadata?). Baseline: 100% of A+ tools document return types.