Integration

FFmpeg Micro integration

FFmpeg Micro runs FFmpeg in the cloud. Transcode, stitch, trim or caption video, audio and image files given by public URL (quality and resolution presets, or allowlisted raw FFmpeg options and a text-overlay option), transcribe audio to SRT subtitles with Whisper, run ready-made blueprints (captions, watermark, reframe, quote card, viral short, product ad), and generate images, speech and image-to-video clips. Finished files are fetched through signed download URLs valid 10 minutes. The direct upload flow (presigned storage URLs, upload confirmation and the free probe of uploaded files) is not exposed: pass inputs by public URL. Billing is in tokens: processing costs 1 token per 10 seconds of input duration, metered per second (failed jobs are not billed), and the image, speech and animation tools charge per image, per character or per second of clip; reads, status checks and download links are free. Authentication is an API key from the FFmpeg Micro dashboard, sent as a Bearer token.

Connect FFmpeg Micro How integrations work

Ask it in chat

Say what should happen.
The workflow builds itself.

When this happens, do that in FFmpeg Micro.Built from the 18 FFmpeg Micro tools below, on the connection you add once.

What you can do with FFmpeg Micro

Each of these is one operation on one endpoint, available both as a workflow step and as a tool an AI agent can call on its own. Nothing bundles several operations into one action, so what a run did stays readable afterwards.

Each badge shows the default agent permission of an endpoint (allowed: an agent calls it without asking; ask first: you are asked before an agent calls it, when the agent or chat asks permission for sensitive actions; never: no agent may call it), which every user can change endpoint by endpoint in their credential settings.

Most endpoints run on the free plan. A few, mainly the ones that publish on your behalf, need a paid plan.

Transcode, 4 operations

cancel_transcodeCancel transcodePATCHAgents: ask firstCancels a transcode job that is still pending or processing; it ends with status canceled. A completed, failed or already canceled job cannot be cancelled (HTTP 400). Returns success, message and the job.
create_transcodeCreate transcodePOSTAgents: ask firstRuns FFmpeg on 1 to 10 input URLs, billed 1 token per 10 s of input; returns the finished job (status, outputs). Get the file with get_transcode_download_url. Waits for the job and returns the finished result; gives up after 10 minutes.
get_transcodeGet transcodeGETAgents: ask firstGets one job by id: status (pending, processing, completed, failed, canceled), inputs, outputs, billable_minutes, timings, lane and error_message. For the file itself call get_transcode_download_url.
list_transcodesList transcodesGETAgents: ask firstLists the account's jobs, newest first by default, as {items, page, limit, total}; transcribe jobs appear here too (job_type tells them apart). Filter by status and by a time window on updated_at.

Blueprint run, 3 operations

continue_blueprint_runContinue blueprint runPOSTAgents: ask firstResumes a caption-video run paused at awaiting_review with the reviewed SRT text; returns the finished run with output_url (signed, valid 10 minutes). Waits for the job and returns the finished result; gives up after 10 minutes.
get_blueprint_runGet blueprint runGETAgents: ask firstGets a blueprint run: status (pending, processing, awaiting_review, completed, failed), current step, output_url (signed, valid 10 minutes, once completed), error_message, tokens_spent and srt_text while awaiting review.
start_blueprint_runStart blueprint runPOSTAgents: ask firstRuns a recipe (slug) on URLs: product-ad costs 200 tokens, the rest are metered as transcodes. caption-video stops at awaiting_review: finish with continue_blueprint_run. Waits for the job and returns the finished result; gives up after 10 minutes.

Transcription, 3 operations

create_transcriptionCreate transcriptionPOSTAgents: ask firstTranscribes speech of an audio or video URL to SRT (Whisper), 1 token per 10 s of input; returns the finished job. Get the SRT with get_transcription_download_url. Waits for the job and returns the finished result; gives up after 10 minutes.
get_transcriptionGet transcriptionGETAgents: ask firstGets one transcription job by id: status (pending, processing, completed, failed, canceled), media_url, duration, billable_minutes, output_url (gs:// SRT once completed) and error_message.
list_transcriptionsList transcriptionsGETAgents: ask firstLists the account's transcription jobs only, newest first by default, as {items, page, limit, total}. Filter by status and by a time window on updated_at.

Animation, 2 operations

animate_imageAnimate imagePOSTAgents: ask firstAnimates a still image into a 5 or 10 s video clip, charging 18 tokens per second of clip (refunded on failure); returns the render with its video url. Waits for the job and returns the finished result; gives up after 10 minutes.
get_animationGet animationGETAgents: ask firstGets an image-to-video render: status (queued, processing, completed, failed), the signed video url once completed (download promptly), detail when failed (tokens refunded) and tokens_charged.

Image, 1 operation

generate_imageGenerate imagePOSTAgents: ask firstGenerates one image from a text prompt and returns its url. Charges tokens: 6 with flux-schnell (used when left out), 10 with nano-banana or with reference_urls. Limited to 30 calls per minute.

Narration translation, 1 operation

translate_narrationsTranslate narrationsPOSTAgents: ask firstTranslates an ordered list of narrations into another language, keeping their count and order, so images made for the source language can be reused. Free (0 tokens); limited to 60 calls per minute per API key.

Scene, 1 operation

split_script_into_scenesSplit script into scenesPOSTAgents: ask firstSplits a narration script (any language) into ordered scenes, each with its narration text and an English image prompt. Free (0 tokens); limited to 60 calls per minute per API key.

Speech, 1 operation

generate_speechGenerate speechPOSTAgents: ask firstReads text aloud with a narration voice (ElevenLabs multilingual, 29 languages) and returns the mp3 url. Charges 25 tokens per 1,000 characters of text, pro rata. Limited to 30 calls per minute.

Transcode download url, 1 operation

get_transcode_download_urlGet transcode download urlGETAgents: ask firstReturns a fresh signed HTTPS URL of a completed transcode's output file, valid 10 minutes (the job's output_url is a gs:// address no HTTP client can fetch). The job must be completed, else HTTP 400.

Transcription download url, 1 operation

get_transcription_download_urlGet transcription download urlGETAgents: ask firstReturns a fresh signed HTTPS URL of a completed transcription's SRT file, valid 10 minutes. The job must be completed, else HTTP 400.

Use FFmpeg Micro from Claude Code, Cursor or Codex

Connect FFmpeg Micro in LiveContext, create an API key in Settings, then add the LiveContext MCP server to your client. Its FFmpeg Micro operations become tools the client can call, next to every other integration you connected.

Claude Code

claude mcp add --transport http livecontext https://livecontext.ai/mcp --header "X-API-Key: YOUR_API_KEY"

Codex (~/.codex/config.toml)

[mcp_servers.livecontext]
url = "https://livecontext.ai/mcp"

[mcp_servers.livecontext.http_headers]
"X-API-Key" = "YOUR_API_KEY"

Cursor, Claude Desktop and other JSON configs

{
  "mcpServers": {
    "livecontext": {
      "type": "http",
      "url": "https://livecontext.ai/mcp",
      "headers": {
        "X-API-Key": "YOUR_API_KEY"
      }
    }
  }
}

Set up the MCP server step by step

How to connect FFmpeg Micro

Create your workspace
Sign up free. Nothing to install.
Connect FFmpeg Micro once
Add your FFmpeg Micro connection in Settings. Every tool from this integration reuses it, and secrets are encrypted at rest and visible only to you.
Describe the job in chat
The workflow builds itself in front of you and picks the endpoints it needs. Drag the steps yourself on the canvas whenever you would rather be explicit.