Developers

Transcription in your stack, or in your AI

Two ways in: a REST API with exactly two endpoints, or a remote MCP server your AI assistant connects to in one paste. Both run the same engine as the site: 95%+ accuracy on clear audio, 90+ languages, speaker labels. Included with any subscription.

Get your API key Connect the MCP server

The API: two endpoints, nothing to learn

Create a token in your account, start a job, poll until it is done. The transcript comes back as plain text with the same accuracy the editor shows.

1. Start a transcription

curl -X POST https://api.filetotext.ai/v1/api/transcribe \
  -H "Authorization: Bearer <your-token>" \
  -H "Content-Type: application/json" \
  -d '{ "url": "https://example.com/interview.mp3", "verbatim": false }'

2. Poll until state is done, read txt

curl https://api.filetotext.ai/v1/api/transcription/<id> \
  -H "Authorization: Bearer <your-token>"

Bearer tokens are created in your account, where the full reference lives. Jobs take a direct media URL (https, pointing straight at the file), with an optional verbatim mode that keeps every filler word.

The MCP server: transcription inside your assistant

FileToText runs a remote MCP (Model Context Protocol) server. Point your AI client at one URL, sign in with your FileToText account when it asks, and your assistant can transcribe any media link mid-conversation. No keys to paste, no local install.

Server URL

https://api.filetotext.ai/mcp

Claude

Settings, then Connectors, then Add custom connector. Paste the URL.

ChatGPT

Settings, then Connectors, then New connector. Paste the URL, pick OAuth. Needs a paid ChatGPT plan.

Cursor

Settings, then MCP, then Add new server. Paste the URL, pick OAuth.

Cline

Open the MCP servers panel, add a remote server, paste the URL.

Then just ask

Transcribe this: <url>
Transcribe this verbatim: <url>

Cleaned-up text by default (ums and uhs removed), verbatim on request. Long files keep processing in the background and your assistant picks up the result.

Access and pricing

API and MCP access are included with any subscription; usage draws from your plan's monthly minutes, with no separate API pricing to model. Start with a free account (first 10 minutes of any file free), then see plans for included minutes. Shipping transcription inside your own product, or expecting serious hours through the API? Get your volume price, personal price by email within 4 business hours.