Loading ToolBench
MCP server that exposes developer identity context to AI agents — memories, active projects, and developer preferences learned from Claude Code
The MCP server exposes 3 tools with significant definition gaps. Only get_developer_context and get_active_project are visible in the MCP package source (packages/mcp/src/index.ts); add_memory is inferred from apps/server code. Tool descriptions exist but are vague (e.g., 'Returns what Ō knows about the developer' lacks WHEN/WHY context). Input schemas for MCP tools are not visible in the provided source, only the Anthropic tool definitions in apps/server are shown. No output schemas documented. Parameters lack type constraints and validation guidance. Error handling is absent. The server conflates two separate concerns: an MCP interface (packages/mcp) and a Fastify backend (apps/server), but the MCP tool definitions themselves are incomplete.
Save something important about the developer for future sessions. Use this when you learn preferences, constraints, facts about their stack, or project-specific context worth remembering.
Returns the currently active project (updated in last 24h), its open tasks, and associated files. Returns empty if no project is active.
Returns what Ō knows about the developer — memories ordered by importance. Call this at the start of a session to understand who you're working with.
Input schemas not visible for get_developer_context and get_active_project in MCP package source. Only Anthropic tool definitions shown in apps/server. Cannot verify JSON Schema compliance for MCP tools.
No output schemas documented for any tool. LLMs cannot plan downstream calls or extract required fields (e.g., does get_active_project return project_id, tasks array, file paths?). Violates pattern:tool requirement.
Descriptions lack WHEN/WHY context. 'Returns what Ō knows about the developer' does not explain when to call this vs other tools, what format memories are in, or how many results to expect. Baseline: descriptions should be 50-200 chars with clear intent.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 41 | 2026-07-28+ | v2 |
add_memory schema visible in Anthropic tools (apps/server) but not in MCP package. Importance parameter lacks min/max constraints in description (should state '1-10 range'). Category enum is present but description does not explain when to use each category.
No error handling guidance. What happens if developer has no memories? If project lookup fails? If memory storage is full? LLMs receive no recovery hints.