Local MCP memory/workspace server with SQLite-based indexing, full-text search, and vector search capabilities
LocalClaw has 7 tools with basic naming and descriptions, but significant gaps in schema completeness and parameter documentation. Tool names follow verb_noun convention (memory_create, memory_search, workspace_list), which is good. However, 3 workspace tools (status, list, read) are defined in test files rather than production code, making their definitions inferred rather than explicitly registered. Descriptions exist but are terse (10-50 chars), below the 50-200 char baseline for LLM optimization. Input schemas are present but lack required field declarations, type constraints (enums), and detailed parameter descriptions. No output schemas are documented. Error handling is minimal, no recovery guidance or actionable error messages visible in the code.
Create a new memory entry with title and content
Retrieve a specific memory entry by ID
Search memory entries using grep-style pattern matching
Search memory entries using full-text search or vector search
List files and directories in the workspace
Read the contents of a file in the workspace
Get the status of the workspace
Workspace tools (status, list, read) defined in test file (mcp_test.go), not production code. Tool definitions are inferred rather than explicitly registered with ToolRegistration struct.
Descriptions are terse (10-50 chars), below 50-200 char baseline. Examples: 'Get the status of the workspace' (35 chars), 'List files and directories in the workspace' (43 chars). Lack context for LLM selection and when to use vs similar tools.
Input schemas lack 'required' field declarations and enum constraints. memory_search 'mode' parameter accepts 'fts' or 'vector' but no enum constraint visible. memory_grep has 8 parameters with no indication which are required.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 48 | 2025-06-18+ | v2 |
No output schemas documented. LLMs cannot infer what fields memory_search returns, what structure workspace_list produces, or how to chain results to downstream tools.
No error handling guidance visible. Tools return errors but no recovery hints (e.g., 'If memory not found, try search_memory() first'). No categorization of retryable vs fatal errors.