Loading ToolBench
Codex plugin and MCP server for Fathom AI Notetaker.
Fathom-MCP has well-structured tool definitions with consistent naming (verb_noun pattern) and comprehensive parameter schemas using Zod. All 11 tools have descriptions and input schemas. However, descriptions are often terse (many under 100 chars), parameter descriptions lack actionable constraints (e.g., no guidance on date formats, pagination limits, or error recovery), and output schemas are not documented. Error handling is minimal, tools return raw API responses without guidance on what to do if a call fails. No tool annotations (readOnlyHint/destructiveHint) despite clear risk levels. The server follows basic MCP patterns but lacks production-grade polish around error messaging and output documentation.
Create a Fathom webhook for new meeting content. At least one include_* option must be true.
Delete a Fathom webhook by ID.
Show which configured Fathom account is used when no account is specified.
Fetch both summary and transcript for a Fathom recording and return them together.
Get the summary for a Fathom recording, or send it asynchronously to a destination URL.
Get the transcript for a Fathom recording, or send it asynchronously to a destination URL.
No output schemas documented. Tools return raw API responses (jsonText wrapper) without specifying what fields the LLM should expect. This forces agents to guess at response structure and breaks tool chaining when downstream tools need specific IDs or references.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint). Tools are marked with Risk levels in comments but not exposed via MCP protocol. fathom_delete_webhook is DESTRUCTIVE and fathom_create_webhook is WRITE, but LLMs cannot see these hints to adjust planning.
Parameter descriptions lack actionable constraints. 'created_after' and 'created_before' in fathom_list_meetings have no format guidance (ISO 8601? Unix timestamp?). 'cursor' pagination parameter has no explanation of how to use it or when to stop. LLMs cannot infer these details.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 59 | 2026-07-28+ | v2 |
List configured Fathom accounts without exposing API keys.
List meetings recorded by you or shared with your team. Optionally include transcript, summary, action items, and CRM matches.
List Fathom team members, optionally filtered by team name.
List teams visible to the authenticated Fathom API key.
Verify Fathom webhook headers and raw request body using FATHOM_WEBHOOK_SECRET or an explicit secret.
No error handling or recovery guidance. Tools return raw API errors without context. If fathom_list_meetings fails with a 401, the LLM has no guidance on whether to retry, check credentials, or try a different account parameter.
Pagination not fully documented. fathom_list_meetings, fathom_list_teams, and fathom_list_team_members accept 'cursor' but descriptions do not explain: Is cursor required? What does it return? How does the LLM know when to stop paginating? No total count or next_cursor field documented.