Five tools with complete input schemas and descriptions, but quality is uneven. Tool names follow verb_noun convention (ingest_history, query_history, ask_history, stats_history, backfill_occurrences). Descriptions are present but brief (avg 95 chars, below 194-char baseline). Parameters have types and descriptions, but lack enums for constrained inputs (e.g., 'browser' accepts 'safari|chrome|firefox|edge|brave|all' but no enum constraint). Output schemas are not documented, responses return dicts with generic 'ok'/'rows'/'stats' keys but no field-level documentation. Error handling is absent; no guidance on retryability or recovery. ask_history uses heuristic NLP parsing without explaining limitations or fallback behavior.

• Add enum constraints to 'browser' (safari, chrome, firefox, edge, brave, all) and 'category' (ai, github_project, personal_site, docs, news, social, shopping, video, other) parameters. Replace free-form strings with formal constraints.
• Document output schemas for all tools. E.g., query_history returns {'rows': [{'url': str, 'title': str, 'category': str, 'visit_count': int, ...}], 'total': int}. Include field types and meanings so LLMs know what to extract.
• Expand tool descriptions to 100 - 200 chars and include WHEN to use each tool. E.g., 'Query browser history with filters (category, domain, time range). Use this for structured searches; use ask_history for natural language questions.'
• Add error handling and recovery guidance. Document what happens on failure (e.g., 'If ingest fails due to missing browser, try specifying browser=chrome explicitly'). Return structured error responses with actionable next steps.
• Document ask_history limitations and fallback behavior. Explain that it uses keyword matching and may not understand complex questions. Suggest query_history as a fallback for precise filtering.