Single tool 'append_transcript' has a clear verb-noun name and documented schema with typed parameters. However, the description (89 chars) is below the 194-char baseline and lacks critical context: WHEN to use it, what happens on retry, whether it's idempotent, and what the response contains. Parameter descriptions are present but minimal. No output schema is documented. The tool modifies state (WRITE risk) but the description does not explicitly warn about side effects or idempotency. No error handling guidance provided.

• Expand tool description to 150 - 250 chars: explain what 'raw' vs 'clarified' means, state that the operation is idempotent on (sessionId, timestamp), and describe the response (e.g., 'Returns {entryId, timestamp, kind}').
• Document the output schema explicitly in the tool definition or description. Include fields like entryId, timestamp, sessionId, kind, and any metadata.
• Add error handling guidance: 'If sessionId is invalid, call search_sessions() first. If MongoDB is unavailable, retry after 5 seconds.'
• Clarify the 'kind' enum: 'raw = unprocessed user input; clarified = LLM-refined version. Choose based on whether the text has been edited.'
• Consider adding a dry_run parameter (boolean, default false) to preview the append without persisting, reducing accidental duplicates.