compute queuesPrivate job queues you can add compute to

A queue is a URL. A job is a Docker container. Push work from a browser or a backend with no authentication and no setup, then add compute by running one command on any machine you own.

Quickstart

Backend integration

Open the web client

GitHub

Browser Docker client

A web page that configures and runs a container, streams its logs, and hands back its outputs. Embeddable in any site as an iframe.

Bring your own compute

One `docker run` on a laptop, workstation, or cluster node joins that machine to your queue. Scale by starting more.

Build it here, not locally

Ship a Dockerfile or a git repo and the worker builds it. Separate build and run logs tell you which half broke.

Three ways to get results

Poll, subscribe over a websocket, or register a callback. Same job, pick whichever fits your backend.

Built for agents

An Agent Skill and an llms.txt so an AI can build a container, debug it, and drive a queue correctly on the first try.

30 seconds

# 1. pick any queue name — unguessable means private
QUEUE=my-$(uuidgen | tr 'A-Z' 'a-z')

# 2. add compute: this machine now works that queue
docker run --rm -v /var/run/docker.sock:/var/run/docker.sock -v /tmp:/tmp \
  metapage/metaframe-docker-worker:0.54.80 run --cpus=2 $QUEUE

# 3. push a job from anywhere
curl -s -X POST https://container.mtfm.io/q/$QUEUE \
  -H 'content-type: application/json' \
  -d '{"definition":{"image":"alpine:3.19.1","command":"sh -c \"echo hi > /outputs/out.txt\""}}'