Open source. Agent directed.

Give the camera
to your agent.

AgentTake turns any AI agent into a video director. Point it at a web app — it drives a real browser, decides what to click, narrates what matters, moves the camera, and walks away with a polished, narrated MP4. No human in the loop, no editing suite, no screen-recording plugins.

Get it on GitHub See how it works

MIT licensed · Free to run · Bring your own agent · Formerly Reel-studio.

agenttake / agenthotel-launch.mp4AGENT DIRECTED

Not a hypothetical.

Directed by an agent, end to end: a Hermes agent running on AgentHotel made AgentHotel's launch video with AgentTake.

The director is already on your team

Your agent. A whole production studio.

It's not another autopilot. It's a clean MCP surface that hands the camera to your agent. The agent that already knows your product is the best director you'll ever hire.

Agent-directed capture

observe returns a live screenshot plus clickable elements; act clicks, types, scrolls, hovers and narrates.

A real camera

Shots push in on what the voice names (wide, medium, close), ease at the speed the director sets, and follow clicks. Recorded at high pixel density so a close-up stays sharp.

Narration on the beat

Speech placed on the exact moment its content is on screen. Free Edge voices by default, ElevenLabs optional; the agent can cast its own narrator.

No dead air

The time an agent spends thinking between steps is cut; long waits can play as a labelled time-lapse.

The director sees

Preview the exact frame before a beat is recorded, and review the finished film as a storyboard contact sheet.

Edit without a reshoot

Hide a step, hold a result, speed up a wait or rewrite a line, then re-render. No re-recording.

Any format

16:9 for X and YouTube, or 4:5 from a site's phone layout for the LinkedIn feed.

Branded and ready to ship

Intro and closing cards built from the video's own frames, a call to action with a QR code, smooth transitions, optional music bed.

Secrets stay secret

Masks blur keys and passwords from the first frame, and passwords are typed by AgentTake itself, never passed through the directing model.

Publishes itself

Every finished film gets a public watch page with its storyboard.

From prompt to premiere

Three steps. Then, action.

Run AgentTake

Start the studio with Docker. Your server, your camera.

Connect your agent over MCP

Claude, Hermes, OpenClaw, Cursor — any MCP client.

Ask for a film

“Make a 45-second tutorial of our signup flow.”

mcp-config.json

{
  "mcpServers": {
    "agenttake": {
      "url": "https://reel.example.com/mcp",
      "headers": {
        "Authorization": "Bearer <REEL_API_TOKEN>"
      }
    }
  }
}

Your infrastructure. Your production.

Run it your way.

One Docker container — Chromium, Xvfb and ffmpeg behind an MCP server on port 8000. Needs 2 CPU cores and 2 GB RAM. Runs anywhere Docker runs.

Ask your agent to install it

Paste this to an agent that has access to your server.

prompt

Install AgentTake from https://github.com/magnusfroste/agenttake on this server with Docker Compose, following its README. Use the domain reel.example.com with HTTPS, generate a long random REEL_API_TOKEN, and then add it to your own MCP servers so you can direct videos with it.

Docker on your own server

Clone, set two variables, start.

terminal

git clone https://github.com/magnusfroste/agenttake.git && cd agenttake
cp .env.example .env   # set REEL_API_TOKEN and REEL_PUBLIC_BASE_URL
docker compose up -d --build

Easypanel, Coolify and similar

Create an app from the GitHub repo, set REEL_API_TOKEN and REEL_PUBLIC_BASE_URL, mount a volume at /data, expose port 8000.

AgentHotel

A ready template, installed as Take: pick it, give it a domain, deploy. The token is generated for you.

Explore AgentHotel

Give the camerato your agent.