Skip to MCP reference

MCP documentation

Your transcripts, inside your AI assistant.

Connect your clientExplore the tools
On this page

Taption MCP

Connect Taption to an MCP-compatible client to transcribe and translate audio and video, read completed transcripts and translations, and browse your recent work without leaving the conversation.

The Model Context Protocol (MCP) lets your assistant call Taption tools on your behalf. This custom connection supports reading transcripts and starting new transcription and translation jobs.

Streamable HTTPOAuth 2.1

Server URL

https://api.taption.com/mcp-custom

Use this full URL when adding Taption as a remote MCP server.

Transport
Streamable HTTP
Authentication
OAuth 2.1

Requirements

  • A Taption account. No paid subscription is required. If you sign in with email and password, verify your email address before connecting.
  • Available minutes to create transcription and translation jobs. Your account’s plan and usage limits apply.
  • A client that supports remote Streamable HTTP servers and OAuth sign-in. Client plans and workspace policies may affect access.

Supported media sources

Use a direct audio or video URL, a public Google Drive file, or a YouTube video. The custom endpoint also accepts public posts from Facebook, Instagram, TikTok, X, Threads, Vimeo, Dailymotion, Bilibili (bilibili.tv), and Twitch videos or clips.

ChatGPT can also supply an uploaded audio or video file when its file integration is available. Other clients can use a downloadable media URL.

Links must be accessible to Taption. Live streams are not supported; use the finished recording.

Configure your client

Choose your application for a complete, step-by-step setup guide.

Authentication

Taption uses OAuth 2.1 with PKCE. Sign in to your Taption account and approve the connection. Your client handles authorization and refreshes the one-hour access token automatically.

MCP uses OAuth access tokens. Taption REST API keys cannot authenticate this connection.

Keep the /mcp-custom URL throughout setup. A token issued for /mcp cannot be used at /mcp-custom.

To revoke access, open Account → Connected apps in Taption and disconnect the client. Reconnect and authorize again when you want to restore access.

Connection permissions

transcripts:read
Read transcript status and content, and list recent transcripts.
transcripts:write
Create transcription and translation jobs using your account’s minutes.

Supported workflows

Give your assistant the media and the intended output. It uses Taption’s tools to complete these steps.

  1. Choose the media and output

    Provide a link or supported upload, the spoken language, and a purpose: a plain transcript, speaker labels, or captions. The assistant asks for missing choices before starting.

  2. Transcribe and follow progress

    transcribe returns an ID immediately. get_transcript waits briefly for a job that is about to finish; if it is still processing, ask again in about a minute. If email notifications are on in Taption, you also get an email when it is done.

  3. Read, summarize, and export

    Read the completed text in your assistant or ask it to summarize the content. Follow the SRT download and editor links when available. The Taption editor offers additional export formats.

Try asking your assistant

Transcribe this recording in English with speaker labels for two speakers: [your media URL].
Create captions for this Japanese video: [your media URL]. Use 16 characters per caption line.
Transcribe this English interview as captions, then translate it into Japanese and Spanish: [your media URL].
Translate my “Weekly sync” transcript into Traditional Chinese.
Find my most recent Taption transcripts and summarize the completed interview. Include the subtitle download and editor links if available.

Replace [your media URL] with your own link. Summaries are produced by your assistant from the transcript; caption length customization depends on your Taption plan.

transcribe

Creates a jobtranscripts:write

Start a background transcription job. Provide either media_url or file. The spoken language and purpose must come from the user; omitting either prompts for the missing choice instead of starting a job. Add translate_to to translate the transcript automatically once transcription finishes.

Parameters

media_urlstringOne media source

A supported media link, up to 2,048 characters. Supply this or file.

fileobjectOne media source

A client-supplied upload with download_url and file_id; mime_type and file_name are optional. The URL must be downloadable by Taption.

languagestringNeeded to start

Spoken language tag, such as en-US, zh-TW, ja-JP, or es-ES. Ask the user; do not infer it from the file name or chat language.

View supported language tags
purposestringNeeded to start

transcript: punctuated text; speakers: speaker labels; captions: short subtitle segments. A request to “transcribe” alone does not select a purpose.

transcriptspeakerscaptions
speaker_countinteger

Estimated speaker count for speakers: 2–10. Defaults to 2 when no estimate is provided.

max_charactersinteger

Characters per caption line: 1–200, for captions on eligible paid plans. Defaults to 16 for zh/ja/ko/yue and 65 otherwise.

translate_tostring[]

Optional. 1–5 languages to translate into once transcription finishes, such as en, ja, or zh-TW. Each uses minutes equal to the media length, in addition to transcription.

extensionstring

File extension for an uploaded file when it cannot be inferred. Not needed for links. Audio and supported video formats are accepted; 3gp is unsupported.

namestring

Transcript name, up to 200 characters. Uses the source file name when omitted.

Example tool arguments

{
  "media_url": "https://example.com/interview.mp4",
  "language": "en-US",
  "purpose": "speakers",
  "speaker_count": 2,
  "translate_to": [
    "ja"
  ],
  "name": "Interview"
}

Illustrative values: replace the media URL or transcript ID with your own.

Returns the transcript ID and status. Each successful creation starts a new job and consumes transcription minutes, plus minutes equal to the media length for each translate_to language; re-use the ID to check progress instead of submitting again.

get_transcript

Read onlytranscripts:read

Check progress or read a completed transcript as timestamped text, SRT, or VTT. Pass language to read one of its translations. The first page includes download_url and editing_url when available, and lists the transcript’s translations and their status.

Parameters

idstringRequired

Transcript ID from transcribe or list_transcripts. 1–128 letters, numbers, underscores, or hyphens.

languagestring

Translation to read, as its language code, such as ja. Omit to read the original transcript.

formatstring

Output format. Defaults to text.

textsrtvtt
offsetinteger

Zero-based phrase offset. Defaults to 0. Use the returned next_offset for the next page.

limitinteger

Maximum phrases per page: 1–500. Defaults to 200.

Example tool arguments

{
  "id": "TRANSCRIPT_ID",
  "format": "text",
  "offset": 0,
  "limit": 200
}

Illustrative values: replace the media URL or transcript ID with your own.

Pass next_offset back as offset to read the next page. Pages contain at most 30,000 characters; only the first 2,000 phrases are readable through MCP. Use Taption for the full transcript. A completed recording without speech returns total_phrases: 0.

translate_transcript

Creates a jobtranscripts:write

Translate an existing, completed transcript into up to five languages. If you refer to a transcript by name, the assistant finds it with list_transcripts and asks you to choose when it is not a clear match. The target languages must come from the user.

Parameters

idstringRequired

Transcript ID from transcribe or list_transcripts. 1–128 letters, numbers, underscores, or hyphens.

target_languagesstring[]Required

1–5 translation language codes, such as en, ja, es, or zh-TW. A transcript cannot be translated into its own language.

overwriteboolean

Set to true to replace an existing translation, including edits made in Taption. Only after the user agrees.

Example tool arguments

{
  "id": "TRANSCRIPT_ID",
  "target_languages": [
    "ja",
    "es"
  ]
}

Illustrative values: replace the media URL or transcript ID with your own.

Returns the languages queued and the minutes each uses, equal to the media length per language. An existing translation is replaced only with overwrite, after you agree, because it may contain your edits. Read the result with get_transcript and language.

list_transcripts

Read onlytranscripts:read

Browse your most recently modified transcripts, newest first. Returns IDs, names, statuses, languages, translations, and modification times. Pass query to find a transcript by name. Use an ID with get_transcript to read the content.

Parameters

limitinteger

Maximum transcripts to return: 1–50. Defaults to 10.

querystring

Part of a transcript name to find among your 100 most recent transcripts. Case-insensitive.

Example tool arguments

{
  "query": "weekly sync",
  "limit": 10
}

Illustrative values: replace the media URL or transcript ID with your own.

Returns an empty transcripts array when there are no recordings. This tool has a result limit, not page-based pagination.

Troubleshooting

Connection or permission error

Reconnect with OAuth and allow the requested permissions. Verify that the configured URL ends in /mcp-custom.

The media cannot be downloaded

Check that the file or post is public, supported, and no longer live. For Google Drive, enable link access.

Not enough minutes or too many jobs

Check your available minutes in Taption. For rate or concurrency limits, wait for active jobs to finish and retry after the indicated delay.

Still processing or transcription failed

For an in-progress status, keep checking the same ID about once a minute. For status error, read the failure message and fix the cause before starting another job.

More export formats

Open the transcript in Taption to export TXT, PDF, SRT, VTT, Final Cut Pro XML, JSON, or video with burned-in subtitles. There is no separate MCP export tool.

Read the REST API documentation