Model · skills & subagents · harness

Share your agent config. Anyone can run it.

Publish one manifest. Anyone can fetch it and run it on a compatible machine.

Install oloop

curl -fsSL https://oloop.run/install.sh | sh

Run this in a terminal on your computer. Install details · Inspect script.

Node.js 20+, curl and tar.

After installation: add oloop to PATH

Follow the installer's printed PATH instruction.

export PATH="$HOME/.oloop/bin:$PATH"

Default installation, current shell. A custom OLOOP_HOME needs its printed path instead.

First install oloop and add it to PATH. Use the name and version you received.

Template

oloop pull <owner>/<name>@<version>

Run it from the intended directory. Fetching writes files; it does not run the agent. CLI pulls need a free handle.

Setups are filtered by OS, chip, and memory. Each setup lists its pins, its hardware, and its evidence.

1 Install oloop

Installs with one command

2 Pick a setup for your machine

oloop host start

Picks the local agent setup that matches your OS, chip, and memory.

3 Run a local model

Default runtime root: <project>/.oloop/

Model + local host

4 Harness opens

Pinned harness

outerloop.yaml · AGENT.md · skills/

5 Talk with your agent

You ↔ LLM