Tools, resources and prompts
Read-only tools are annotated readOnlyHint: true. Write tools are annotated readOnlyHint: false and use a
two-step confirmation (see below).
| Tool | What it does | Permission |
|---|---|---|
search_meetings | list or search meetings (title, report and transcript text), filter by client, tags, meeting date, status | read:meetings |
get_meeting | metadata, participants, AI report and documents of a meeting | read:meetings |
get_meeting_summary | the AI meeting report (Markdown and structured) | read:meetings |
get_participants | participants and speaker labels | read:meetings |
get_transcript | diarized timestamped transcript, split in parts (part, max_tokens_per_part) | read:transcripts |
search_transcripts | find what was said: excerpts with timestamps and speakers across all meetings | read:meetings + read:transcripts |
export_meeting_for_llm | the LLM-ready bundle of a meeting, in parts | read:meetings + read:transcripts |
list_personas, list_templates | professional personas and document templates | read:meetings or write:documents |
generate_document | generate a document from a template or a custom prompt (job, billed) | write:documents |
list_documents, get_document | generated documents and their Markdown content | read:documents |
create_meeting | create a meeting from a public https audio or video URL, with optional documents (job, billed) | write:meetings |
update_meeting | rename a meeting, change its date or tags, add participants and say which speaker is who | write:meetings |
send_feedback | report a bug or request a feature to the Meetingize team | any |
get_job, list_jobs | status of meeting processings and document generations | any |
get_usage | API consumption of the month, wallet and spend cap | read:usage |
get_account | connected account, plan and granted permissions | any |
documentation | the whole Meetingize documentation (index, a page, or everything) | none |
search, fetch | ChatGPT search and deep research contract | read:meetings or read:documents |
Tool results contain readable text for the model and structuredContent (JSON) for clients that use it. Errors
(missing permission, not found, quota) come back as tool results with isError: true and an explanation the
assistant can relay.
Long transcripts
Section titled “Long transcripts”get_transcript returns part 1 of N with about 12,000 tokens per part by default, never cutting a sentence, and
tells the assistant how to fetch the next part. This keeps two-hour meetings usable in any context window.
Search
Section titled “Search”search_meetings and search_transcripts use case-insensitive full-text matching on titles, reports and
transcripts. Semantic search is Coming soon.
Two-step confirmation for write tools
Section titled “Two-step confirmation for write tools”generate_document and create_meeting consume your API allowance; update_meeting changes a meeting and
send_feedback sends a message. Called without confirm: true they do nothing and return a preview (“generate a document from meeting X with template Y, billed on your allowance”).
The assistant must show it to you and call again with the same arguments and confirm: true. The confirmed call
carries an idempotency key derived from the arguments, so a retry never starts the same job twice.
Resources
Section titled “Resources”| URI | Content |
|---|---|
meetingize://meetings/{meeting_id} | meeting metadata, participants and AI report (Markdown) |
meetingize://meetings/{meeting_id}/transcript | full transcript (Markdown) |
meetingize://documents/{document_id} | a generated document (Markdown) |
meetingize://docs/{slug} | a documentation page, meetingize://docs/llms-full.txt for all of it |
resources/list returns the documentation pages and your 20 most recent meetings.
Prompts
Section titled “Prompts”| Prompt | Arguments | Workflow |
|---|---|---|
prepare_meeting | who, topic (optional) | briefing before a meeting: history, decisions, commitments, open questions, agenda |
extract_commitments | who (optional), period (optional) | table of commitments made by you and to you, with sources |
followup_email | meeting_id, tone (optional) | follow-up email draft (the meeting report is attached) |
weekly_digest | days (optional, default 7) | digest of recent meetings: decisions, open actions, risks |