Can you use MCP to export meeting transcripts in batches?

Start with the evidence table below. Most major note takers now ship an official MCP server, but plan gates differ. Otter lists its MCP server from the Basic plan, yet bulk export requires Business or Enterprise. Fireflies' free plan has no downloads at all. Granola's MCP can only return transcripts on a paid plan, and its Basic plan is limited to the last 30 days of personal notes. Notta is the least restrictive option here: its MCP server and CLI launched in September 2026 for all users on all plans, with no separate tier to unlock transcript access.

Most vendors expose a remote MCP endpoint you add as a custom connector, then sign in with OAuth. For example, Fathom's documented setup in Claude is: claude mcp add fathom -- npx mcp-remote@latest https://api.fathom.ai/mcp. Notta works differently: its MCP server runs locally, which keeps your transcript traffic on your machine and lets the assistant write exported files directly to a folder you allow. Fellow and Read AI require a workspace admin to enable the connection first, and Claude Team and Enterprise workspaces may need admin approval for custom connectors.

Ask your assistant for a date range, folder or search term, then page through the results. Per-request limits vary by vendor: Fireflies returns up to 50 meetings per call, Read AI up to 10 per page, and tl;dv up to 50 per search. Expect the assistant to make several list calls for anything more than a month of meetings. If your tool supports folders or tags, filtering by them first keeps the batch small and predictable.

MCP returns one transcript per call, so the assistant loops through your list. Ask it to write each transcript to a file named by date and title, for example 2026-10-14-sprint-review.md. Only clients with file access can do this: Claude Code and Notta's local MCP server can write exports to disk, while browser-based clients can only show the text for you to copy. Notta's MCP can export each chosen recording as PDF, Word, text, Excel or SRT, which is handy when the destination expects a specific format rather than raw text.

Compare the number of files against the meeting count in your note taker's app, and spot-check a few transcripts for speaker names and timestamps. Pagination is the most common failure point: some servers skip meetings when filters are set (Plaud's docs note this explicitly), and a long loop can hit rate limits such as Granola's roughly 100 requests per minute. Re-run the list call and diff the two sets of meeting IDs if the counts do not match.

For a whole archive, MCP is the slow route. Otter offers bulk export on Business and Enterprise (Basic exports TXT only). Fireflies provides a full account export on request: a zip with TXT transcripts, summaries, MP3/MP4 audio and video, and JSON metadata, delivered in 24-48 hours with a download link valid for 7 days. Granola exposes an API on its Business plan for scripts and bulk integrations. Notta is the most flexible for repeatable batch exports: its CLI lets you script search, retrieval and downloads, so you can export a folder of transcripts on a schedule using cron or Task Scheduler, in the same PDF, Word, text, Excel and SRT formats the MCP server offers.