Loading ToolBench
Exposes search and read tools for Claude Desktop to query a personal encyclopedia of past Claude.ai conversations stored as markdown summaries with metadata.
Chronicle MCP has well-structured tool definitions with complete schemas and descriptions. All 5 tools are read-only with clear, domain-specific documentation. Naming follows verb_noun convention (chronicle_find, chronicle_read, chronicle_ls). Descriptions are detailed (150-400 chars) and include usage guidance. However, parameter descriptions lack format constraints (e.g., date format, significance enum values not formally declared), and output schemas are not explicitly documented. Error handling is present but generic. No tool annotations (readOnlyHint, etc.) despite all being read-only operations.
Search Alejandro's Chronicle — a personal encyclopedia of his past claude.ai conversations, stored as markdown summaries with metadata. Returns condensed list of top results with date, title, and score. Use chronicle_read on top hits to read full summaries.
List conversations, optionally filtered by period/date-range, project, and/or significance. Provide at least one filter. Two ways to scope by time: period (named label like '2026_Q1') or from_date/to_date (arbitrary day range).
Extract a specific passage from a conversation summary by UUID and optional line range. Returns the extracted text with context.
List all unique project names in Chronicle, with conversation counts per project.
Read the full summary of a conversation by UUID. Returns the complete markdown summary with frontmatter metadata.
Output schemas not documented. Tools return strings or unstructured text; LLMs cannot plan downstream calls or extract structured data reliably.
Parameter constraints not formally declared. 'significance' accepts 'high|medium|low' but no enum constraint in schema; 'from_date'/'to_date' accept YYYY-MM-DD or 'now' but format not specified in schema. LLMs must infer from description text alone.
No tool annotations despite all tools being read-only. readOnlyHint should be set on all 5 tools to signal safety to agents.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 57 | 2026-07-28+ | v2 |
Error messages are generic strings ('No matches for...', 'Search index not found'). No recovery guidance or actionable next steps for LLM.
chronicle_projects has empty input schema (no parameters). Description should clarify it takes no arguments and returns a list of project names with counts.