This Fireflies MCP server provides 4 read-only tools with well-structured schemas and generally good descriptions. All tools are explicitly registered in the TOOLS array with proper JSON Schema definitions including required parameters, types, and descriptions. However, there are notable gaps: output schemas are completely undocumented (critical for agent planning), parameter descriptions lack actionable detail (e.g., date formats mentioned but not validated), and error handling is present but generic. The server follows verb_noun naming conventions well (get_*, search_*, generate_*) and avoids multi-responsibility tools. Tool descriptions range from 80-200 chars (good baseline) but lack dependency hints and recovery guidance. No tool annotations (readOnlyHint, etc.) despite all tools being read-only. The server would benefit from: (1) documented output schemas showing field names/types for downstream tool chaining, (2) richer parameter constraints (e.g., min/max for limit, regex for ISO dates), (3) per-parameter validation with actionable error messages, and (4) tool annotation hints to declare read-only intent to agents.
Generate a summary of a meeting transcript
Retrieve detailed information about a specific transcript. Returns a human-readable formatted transcript with speaker names and text, along with metadata and summary information.
Retrieve a list of meeting transcripts with optional filtering. By default, returns up to 20 most recent transcripts with no date filtering. Note that this operation may take longer for large datasets and might timeout. If a timeout occurs, a minimal set of transcript data will be returned.
Search for transcripts containing specific keywords, with optional date filtering. Returns a human-readable list of matching transcripts with metadata and summary information.
Output schemas completely undocumented. No specification of return fields, types, or structure for any tool. LLMs cannot plan downstream calls or extract needed data.
Parameter constraints lack actionable detail. Date parameters mention 'ISO format (YYYY-MM-DD)' in description but lack formal regex patterns or validation guidance. Limit parameters have no min/max bounds defined (e.g., limit could be 0, 1000000, or negative).
No tool annotations present. All 4 tools are read-only (no state modification), but lack readOnlyHint annotation. Agents cannot infer safety or retry-ability from metadata.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-22 | D | 55 | 2026-07-28+ | v2 |
| 2026-03-09 | F | 0 | - | v1 |
fireflies_generate_summary lacks action context. Description ('Generate a summary of a meeting transcript') is only 54 chars and omits key details: What formats are available? Is this a local summary or API call? When should the agent prefer this over just reading the full transcript?
No pagination guidance in tool descriptions. fireflies_get_transcripts and fireflies_search_transcripts both accept 'limit' but do not document how to fetch additional pages (cursor, offset, etc.). Agents may assume one call fetches all data.
Error messages in executeQuery() are generic ('API request failed', 'GraphQL error'). No recovery guidance for agents. E.g., 401 returns 'Invalid API key' but does not suggest next steps (check config, retry, etc.).
fireflies_get_transcripts timeout handling returns 'minimal set of transcript data' on timeout, but description does not specify what 'minimal' means (just id/title/date?) or how the agent should interpret this fallback.