MCP command reference

Typical async workflow

1. Call `transcribe_file(...)` and keep the returned `job_id`.
2. Then call `get_job(job_id)` to check the status.
3. If status is `queued` or `processing`, wait a bit and call `get_job(job_id)` again.
4. If status is `completed`, read the transcription from `result_text`. If status is `failed`, handle the error payload.

get_usage_summary

Returns the current minute balance and usage summary for the connected account.

No parameters.

{
  "tool": "get_usage_summary",
  "input": {}
}

list_recent_transcriptions

Returns recent transcription jobs with status, file metadata, and text preview.

`limit` optional, from 1 to 50.

{
  "tool": "list_recent_transcriptions",
  "input": {
    "limit": 10
  }
}

get_job

Returns the current status and result payload for a specific transcription job. Use it after `transcribe_file` to poll until the job is `completed` or `failed`.

`job_id` required.

{
  "tool": "get_job",
  "input": {
    "job_id": "job_123"
  }
}

transcribe_file

Uploads a local media file for transcription and creates a new async job. It returns `job_id`, not the final transcript, so call `get_job(job_id)` afterwards to fetch the result.

`file_path` required. `provider`, `language`, and `duration_seconds` optional.

{
  "tool": "transcribe_file",
  "input": {
    "file_path": "/Users/example/audio/founder-call.mp3",
    "provider": "openai",
    "language": "en"
  }
}