Loading ToolBench
Zero-dependency cross-agent conversation history search. Indexes local transcripts of multiple AI coding agents (Claude Code, Codex CLI, OpenCode, Cursor) into a single SQLite database and exposes them via MCP, CLI, or skill.
pawrecall exposes 3 tools with basic descriptions and partial schemas. Tool names follow verb_noun convention (paw_search, paw_read, paw_index), which is good. However, descriptions are terse (50-100 chars), parameter descriptions are minimal, and output schemas are not documented. The paw_index tool accepts no parameters (empty input object), which is unusual and suggests incomplete schema definition. No error handling guidance, no pagination support for search results, and no tool annotations (readOnlyHint/destructiveHint). The server is STDIO-only, which is a hard cap at 50 for protocol readiness.
Trigger an incremental reindex of all agent transcripts. Scans configured agent directories and updates the SQLite database with new or modified messages.
Read the full conversation thread from a specific session. Returns all messages in chronological order for the given session ID.
Search the indexed conversation history by keyword or phrase. Returns matching messages with source, project, session, and file context.
Output schemas not documented. LLMs cannot plan downstream calls or extract fields without knowing what paw_search, paw_read, and paw_index return. Responses are opaque.
paw_search accepts 'limit' parameter but no pagination (offset/cursor). Large result sets will blow context window. No guidance on max results or how to iterate.
Parameter descriptions are minimal or missing. 'source' param in paw_search lists enum values but no description of what filtering by source means. 'project' param has no guidance on format or how to discover valid projects.
paw_index has empty input schema (no parameters). Description says 'Trigger an incremental reindex' but does not explain prerequisites, side effects, or expected duration. No error guidance if reindex fails.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 54 | 2026-07-28+ | v2 |
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). paw_index is a write operation but not marked as such. LLMs cannot distinguish safe reads from state-modifying calls.