Loading ToolBench
MCP server for Meeting BaaS API, allowing AI assistants to manage recordings, transcripts, and calendar data
28 tools with mostly complete schemas and descriptions, but significant gaps in parameter documentation and output schema clarity. Tool names follow verb_noun convention well (joinMeeting, leaveMeeting, searchTranscript). Descriptions average ~100-150 chars, meeting baseline minimums but lacking LLM-optimized guidance on WHEN to use each tool vs. similar ones. Many parameters lack descriptions (e.g., searchTranscriptByType has no visible input schema in source). Output schemas not documented, LLMs cannot predict response structure. Error handling present but generic. Security concern: credentials passed as parameters in setupCalendarOAuth, listRawCalendars, and credential-variant tools violate secret-injection pattern.
Cancel a scheduled recording for a calendar event
Cancel a scheduled recording for a calendar event using direct API credentials
Check the status of calendar integration
Delete a calendar integration
Automatically find and share key moments and topics from a meeting recording with configurable granularity
Find specific topics discussed in a meeting
Credentials exposed as tool parameters (setupCalendarOAuth, listRawCalendars, *WithCredentials variants). API keys, client secrets, and refresh tokens must never appear in tool parameters, they leak into logs and LLM context. Use server-side secret injection via environment variables or vault.
Multiple tools lack visible input schemas in source code (searchTranscriptByType, findMeetingTopic, searchVideoSegment, intelligentSearch). Cannot verify parameter types, descriptions, or constraints.
Output schemas not documented for any tool. LLMs cannot predict response structure, field names, or types. Agents cannot plan downstream tool calls or extract required data reliably. Add explicit response schema documentation to every tool.
Inferred effective spec: <=2025-11-25.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 52 | <=2025-11-25 | v2 |
Get detailed information about a specific calendar integration
Get details of a specific calendar event
Get meeting data including transcripts, video, and metadata
Get meeting data including transcripts, video, and metadata using direct API credentials
Perform intelligent search across meeting content
Have a bot join a meeting now or schedule it for the future. Bot name, image, and entry message will use system defaults if not specified.
Have a bot leave an ongoing meeting
List all calendars integrated with Meeting BaaS
List calendar events with advanced filtering
List calendar events using direct API credentials
List available calendars from Google or Microsoft before integration
List upcoming meetings from a calendar
Get guidance on setting up OAuth for calendar integration
Resynchronize all integrated calendars
Schedule a bot to record a calendar event
Schedule a bot to record a calendar event using direct API credentials
Search within a meeting transcript
Search transcript by content type (questions, decisions, action items, etc.)
Search for specific video segments in a meeting recording
Integrate a calendar using OAuth credentials
Generate a list of links to important moments in a meeting
Generate a shareable link to a specific moment in a meeting recording
Duplicate tools with overlapping functionality (getMeetingData vs getMeetingDataWithCredentials, listEvents vs listEventsWithCredentials, scheduleRecording vs scheduleRecordingWithCredentials). LLMs waste reasoning cycles deciding between them. Consolidate by using server-side auth injection instead of credential parameters.
No tool annotations (readOnlyHint, destructiveHint, idempotentHint) present. LLMs cannot distinguish safe read-only tools from destructive ones (deleteCalendar, leaveMeeting) without explicit hints. Add tool annotations to guide agent planning and prevent accidental data loss.