references/mcp.md
411 tokens · o200k_base · 2,548 bytes
Source excerpt starting at line 1.# MCP tools and resources Cloud MCP is read-only. Local MCP adds proposal tools, which insert or decline staged edits; they never apply those edits to the meeting. | Tool | Use || ------------------------------- | -------------------------------------------------------------------------------------------------------------------------- || `list_meetings` | Find recent meetings by title, ID fragment, or recurring series. An empty list means none matched; do not invent meetings. || `get_meeting` | Read metadata, canonical note, summaries, participants, and action items. || `get_meeting_transcript` | Read a transcript page. Start with `limit: 200`; continue from `pagination.next_offset` only as needed. || `get_recurring_meeting_history` | Find meetings from the same recurring series as a known meeting. || `export_meeting` | Read the complete meeting record including transcripts. Use only when smaller reads are insufficient. || `propose_summary_edit` | Stage a complete summary replacement. Pass `target_id` when multiple summaries exist. || `propose_memo_edit` | Stage a complete memo replacement. || `list_proposals` | List staged proposals. Defaults to `status: pending`. || `get_proposal` | Read one proposal and its unified `diff`. || `decline_proposal` | Discard a pending proposal without changing the meeting. Applied proposals cannot be declined. | Transcript limits are measured in words. The default is 200 and the maximum is 500. Available resources: - `anarlog://meetings/{meeting_id}`- `anarlog://meetings/{meeting_id}/transcript{?offset,limit}`- `anarlog://series/{series_id}` Prefer tools when the workflow needs structured JSON. Use resources when the client needs concise Markdown or plain-text context. Transcript resources default to 200 words (cap 500) and return plain text only.