Open-source agent runtime

Own the agents doing your company's work.

Define agents in Markdown. Run them with Claude, OpenAI, or open models on infrastructure you control.

Let them work autonomously. Put consequential actions behind human approval.

npx -y agentuse@latest run https://agentuse.io/hello.agentuse

One file. Zero babysitting.

One command. Your agent handles the rest.

Trigger your way

0 8 * * *

Runs while you sleep

$ curl -X POST :12233/api/run{"agent":"my-agent"}

HTTP triggers

on: push

GitHub Actions

docker run

Containers

Then close the terminal

The terminal is where you build an agent. It's not where the work lives.

agentuse serve gives every agent a home for autonomous runs, review requests, and results. Same files, same runtime, on infrastructure you control.

01 / Dashboard

One place for every agent: run health, what needs you, and the results they shipped, at a glance.

02 / Agent graph

One system. Not a pile of scripts.

Your agents are a system, not a pile of scripts. Delegation and dependencies, drawn from the files themselves.

03 / Session results

The outcome, not the logs.

Every run ends on the outcome it produced, not a wall of logs.

04 / Mobile approval

Approvals that fit in your pocket.

Approve the calls that touch money or customers. From your desk, or your phone.

✓

Runs unattended: cron jobs, CI/CD pipelines, Docker containers.

✓

Version control just works: diff, review, and merge agents like code.

✓

No SDK required: if you can write plain English, you can build an agent.

✓

Production-ready: retries, streaming, error recovery, and MCP support.

Copilots assist you. AgentUse agents work for you.

AgentUse Studio

Want the operating layer built and running for you?

Founder-led implementation from the team behind AgentUse: recurring functions absorbed one by one, on your own infrastructure.