$5 free credits when you sign up Claim now
Marlin 2b now available Test it!
MiniMax H3 now available! Test it!
Browse all Models
Browse all
Browse all
Browse all

Transcription MCP Server for Claude, Cursor and ChatGPT

Add one MCP server and your agent can transcribe YouTube, TikTok and Twitch links, describe what happens on screen, read text from images and speak. You set the budget on the key, and the agent can check the price before each job.

$5 in free credits to start. No credit card required.

to transcribe 50 twenty-minute YouTube videos
≈ $1.03
mcp.deapi.ai/mcp for every client
1 URL
AI tasks your agent can run, from transcription to speech
+18

MCP Tools for Transcription, Video Description, OCR and TTS

  • Hear

    Transcribe links and files

    YouTube, X, Twitch, Kick and TikTok links, X Spaces, and uploaded audio and video. Word timestamps, speaker labels and TikTok links need model WhisperLargeV3Ct2, which the tool does not pick by default, so name it in your request.

  • See

    Describe video

    A timeline of what happens on screen, with start and end times for each event.

  • Read

    Text from images

    OCR for text in screenshots, slides and scanned documents.

  • Speak

    Text to speech

    Narration from $0.77 per million characters on Kokoro, or in 23 languages on Chatterbox.

  • Remember

    Embeddings

    BGE-M3 vectors (1,024 dimensions) to search what the agent has read and heard.

  • Check

    Price and status

    A price tool next to every job tool and a job status tool, so the agent can quote a job before it spends.

Add the MCP Server to Claude Code, Claude Desktop or Cursor

  1. Create a key for the agent

    In the dashboard, give it a spend limit per day, week or month and an expiry date.

  2. Add the server

    Point your client at https://mcp.deapi.ai/mcp. Claude Code, Claude Desktop and Cursor send the key as a Bearer header; claude.ai and ChatGPT sign in with OAuth.

  3. Ask in plain words

    “Transcribe this video and list every product the speaker mentions.” The agent picks the tool and runs it. Ask it to check the price first if you want a quote.

Setup steps for every client are in the MCP server docs.

What YouTube Transcription Costs Through MCP

deAPI charges for the media task. Your agent's language model is billed separately by its provider.

Task deAPI cost
Transcribe one 20-minute YouTube video $0.021
Research across 50 such videos $1.03
Transcript plus on-screen timeline of a one-hour YouTube video ≈ $0.16
Read a 10,000-character answer aloud on Kokoro $0.0077

From the /price endpoint on 27 October 2026, transcription on model WhisperLargeV3Ct2, timeline on Marlin 2B with the balanced preset. The agent can call the price tools for current amounts.

Spend Limits and Key Controls for Unattended Agents

  • Hard spend limit per key

    Daily, weekly or monthly. Over the limit the API answers 402 key_limit_exceeded and the job does not run.

  • Prepaid credit only

    A key never spends more than the account balance.

  • Expiry and IP allowlist

    Keys can expire on a date and accept calls only from an allowlist of up to 20 IP entries.

  • Price check per tool

    Each job tool has a matching price tool, for example video_url_transcription_price.

  • Webhooks

    Every job tool accepts a webhook URL, so your own systems hear about finished jobs too.

  • Data handling

    Generated files are deleted no later than 24 hours after job creation.

  • $5 in free credits
  • No credit card required
  • Prepaid, no subscription

Give Your Agent a Key with a Budget

$5 in credits after email verification, no card. Setup takes one URL and one key.

See also YouTube Transcript API, Claude Code Skills and deAPI on Apify.

Frequently Asked Questions

Setup, budgets and limits

Create a deAPI key, add the server https://mcp.deapi.ai/mcp to Claude Code, Claude Desktop or claude.ai, and ask Claude to transcribe the link with model WhisperLargeV3Ct2. Claude calls the video transcription tool and returns the transcript.
Yes. The docs have setup steps for Claude Code, Claude Desktop, claude.ai, Cursor, ChatGPT, Codex and Gemini CLI.
From your prepaid deAPI balance, through the API key you give the agent. Set a spend limit on that key. Once it is reached, the API returns 402 and the job does not run.
Yes. Every generation tool has a matching price tool, for example video_url_transcription_price next to video_url_transcription.
Yes. The server is live at https://mcp.deapi.ai/mcp, and its source is on GitHub under the MIT license. The same tasks are also available through the REST API.
No. For thousands of files, use the HTTP API with webhooks.
No. deAPI charges for the media tasks. The language model your agent runs on is billed by its own provider.