mcp-agent-transcriber
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| TRANSCRIBER_MODEL | No | Whisper checkpoint (turbo / large-v3 / medium / ...) | turbo |
| TRANSCRIBER_LANGUAGE | No | Default caption language for direct grabs | en |
| TRANSCRIBER_LOGS_DIR | No | Rotating DEBUG logs (5x5 MB) | logs |
| TRANSCRIBER_OUTPUT_DIR | No | Where transcript .txt/.vtt files go | transcript_output |
| TRANSCRIBER_DOWNLOADS_DIR | No | Where audio downloads land | <output>/downloads |
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {
"listChanged": false
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| get_current_datetimeA | Return the current local and UTC date/time in ISO 8601 format. |
| video_infoA | Resolve a video URL with one yt-dlp metadata pass: platform, title, duration, uploader, available caption tracks, and the audio formats found. Returns JSON; known failures (unsupported URL, private video, sign-in required, network) come back as a status:failed payload. |
| supported_platformsA | List the common tube platforms (YouTube, Vimeo, Dailymotion, Twitch, TikTok, ...) and whether the installed yt-dlp ships an extractor for each. Use before evaluating an unfamiliar host. |
| list_available_transcriptsA | List the caption tracks a video URL exposes: language codes plus whether each is a manual or auto-generated track. Uses yt-dlp metadata only — no download. |
| fetch_transcriptA | Grab an agent-friendly transcript for a video URL without downloading the media. YouTube uses youtube-transcript-api (fast); every other platform falls back to the yt-dlp caption pass. Writes a .txt (plus .vtt on request) into the output dir and returns the cleaned text. |
| download_audioA | Download a video's audio track into the downloads dir via yt-dlp. quality: "best" (default, highest bitrate), "high" (>=128 kbps floor), "standard" (<=128 kbps, smaller), or "low". output_format: "orig" keeps the platform's container; one of mp3/wav/m4a/opus/flac/aac re-encodes with ffmpeg as a postprocess. |
| transcriber_statusA | Report Whisper runtime readiness: model, device (cuda/cpu), whether the checkpoint is already cached/loaded, and ffmpeg availability. Does NOT load the model. |
| transcribe_fileA | Transcribe a local video/audio file with Whisper (turbo). The model loads on first use and stays cached. Writes a .txt (plus .vtt on request) into the output dir and returns the transcript. language: ISO code to force (e.g. "en"); None auto-detects. |
| transcribe_videoA | One-call video → transcript. method: "auto" (default) grabs the platform's captions first and falls back to Whisper only when none exist; "captions" uses captions only; "whisper" downloads the audio (quality: best/high/standard/low) and transcribes locally. keep_audio=true retains the downloaded audio file in the downloads dir; otherwise it is deleted after transcription. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 9 tools
Most tools have distinct purposes: metadata, caption listing, transcript fetching, audio download, Whisper status, and transcription. However, fetch_transcript and transcribe_video (with method 'captions') overlap in functionality, though their descriptions clarify the use cases. The get_current_datetime tool is unrelated but not confusable with others.
Naming mixes verb-first patterns (get_current_datetime, list_available_transcripts, fetch_transcript, download_audio, transcribe_file, transcribe_video) with noun-first patterns (video_info, supported_platforms, transcriber_status). Verbs also vary (get, list, fetch, download, transcribe). The inconsistency is noticeable but not chaotic.
Nine tools is reasonable for a transcription server covering platform checks, metadata, caption handling, audio download, and Whisper integration. The inclusion of get_current_datetime feels out of scope but does not significantly bloat the set. The count is well within the typical range for a focused domain.
The server covers the full transcription workflow: verifying platform support, retrieving video metadata, listing captions, fetching caption transcripts, downloading audio, checking Whisper readiness, transcribing local files, and doing end-to-end video transcription. Minor gaps include no explicit language selection for transcribe_video and no file management, but these are not critical for core use.