MCP server / TikTok

TikTok MCP server: let your AI read TikToks

Paste a TikTok link and your assistant gets the spoken words as text, whether or not the video ever had captions.

Connect it to your client

Any MCP client connects the same way: point it at the server URL and send a TranscriptFetch API key as a bearer header (on claude.ai, OAuth replaces the key). Client-by-client walkthroughs: Claude, Cursor, Codex, or the configuration reference.

Server URL
https://transcriptfetch.com/mcp
Any stdio client (mcp-remote bridge)
npx mcp-remote https://transcriptfetch.com/mcp \
  --header "Authorization: Bearer tf_live_YOUR_KEY"

What it can do

get_transcript accepts a full tiktok.com link or a short share link and returns the words with timestamps. There is no TikTok search or listing tool, links come from you or from wherever your agent found them, but short videos mean small transcripts, so batching a folder of saved TikToks through an agent is quick and cheap.

In your assistant, after connecting
You: What's the hook structure in these five TikToks?
     [five links]

Agent: [get_transcript x5]
       All five open with a question in the first two seconds...

Why TikTok needs transcription, not extraction

TikTok's auto-captions are opt-in per upload, so the majority of videos expose no caption file. A tool that only extracts existing captions comes back empty on most TikTok links, and that is not a bug, there is genuinely nothing to extract. This server transcribes the audio instead, so the caption file's absence does not matter.

The platform's other quirk is burned-in text. A huge share of TikTok captions were typed in the TikTok editor or CapCut and rendered into the pixels; a human reads them effortlessly and no parser sees them at all. Audio transcription sidesteps that too, returning what was said rather than what was overlaid.

What it cannot conjure is speech that never happened: a dance clip or a product montage over a licensed track transcribes to nothing, because nothing was said. Private accounts stay out of reach by design.

Frequently asked questions

Do TikToks even have captions to fetch?

Usually not as files: TikTok's captions are opt-in and most creators skip them or burn text into the frames. The server transcribes the audio with AI, so a missing caption file does not produce a missing transcript.

Can it search TikTok for videos?

No. Discovery tools exist for YouTube only; for TikTok you supply the links. Once your agent has links, it can fetch them in bulk.

What about videos with music only?

No speech means no transcript, and the fetch returns empty rather than inventing text. Failed and empty fetches are never billed.

What does it cost?

100 free credits a month, one credit per successful transcript. TikToks are short, so a credit goes further here than anywhere else.

Short videos, instant text

Sign up, grab a key, paste one config block into your client. Most TikToks transcribe in seconds.

TikTok MCP Server: Video Transcripts for AI Clients