Search Claude Code session history from inside Claude Code via MCP
Server exposes 4 well-named, read-only tools with complete input schemas and descriptions. All tools follow verb_noun naming (search_*, get_*, list_*). Descriptions are present and adequate (60 - 150 chars), exceeding the 10-char minimum. All parameters have type definitions and descriptions. However, output schemas are not formally documented in the code, responses are inferred from return type hints (dict). Parameter descriptions lack detail on constraints (e.g., 'limit' has no min/max guidance; 'days' lacks range). Error handling is minimal, get_resume_command_tool returns {'error': '...'} but provides no recovery guidance. No tool annotations (readOnlyHint, idempotentHint) despite all being read-only. Composition is sound: tools chain naturally (search → get_session → resume). Missing: pagination guidance for large result sets, explicit output schema documentation, and actionable error messages.
Return a copy-pasteable shell command to resume the given session.
Read the full turn-by-turn content of one session.
List sessions started within the last N days.
Full-text search across all indexed Claude Code sessions.
Output schemas not formally documented. Responses are dicts with inferred structure (e.g., search_history_tool returns {query, sessions[...matched_snippets, rank]}). LLMs cannot plan downstream calls without explicit schema documentation.
Parameter constraints missing. 'limit' (default 5) and 'max_turns' (default 200) lack min/max bounds. 'days' and 'since_days' lack range guidance (e.g., 1 - 365). LLMs may pass invalid values.
Error handling lacks recovery guidance. get_resume_command_tool returns {'error': 'session not found'} with no suggestion to search or list sessions first. Errors should guide the LLM to the next action.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 56 | 2026-07-28+ | v2 |
No tool annotations despite all tools being read-only and idempotent. Adding readOnlyHint=true and idempotentHint=true would help agents optimize retry logic and caching.
Pagination not addressed. search_history_tool and get_session_tool accept 'limit' but do not document total count or next_cursor. Large result sets may exceed context window.