Getting started

Harnesser gives an AI agent harness its own machine: a dedicated microVM with persistent storage, a web UI on its own URL, and model access already wired up. You pick a harness, pick a size, and open your workspace a few minutes later.

1. Create an account

Sign up at console.harnesser.co with an email address and a password of at least 8 characters. The console is where you create, monitor, and delete harnesses.

Each account can run up to 3 harnesses at a time.

2. Choose a harness

The create wizard starts with the harness, the software your agents run in:

• Paperclip is an autonomous agent company: issues, agents, and a board that drives them. It ships with the Claude Code, Codex, and OpenCode CLIs.
• QM is a multi-agent chat workbench with skills, runs, and a first-party web UI, including its own accounts for your team.
• Hermes is the Nous Research agent with its full-featured web UI: conversations, memory, scheduled jobs, and a persistent workspace.

Not sure which one fits? See Choosing your harness.

3. Choose a size

Pick the machine your agents will run on. Every size is a dedicated microVM with its own kernel, dedicated memory, and a persistent NVMe disk. See Machine sizes for the full comparison and guidance on which size fits which workload.

The size is fixed when the harness is created. To move to a different size today, create a new harness at the new size.

4. Name your harness

The name becomes your subdomain: a harness named my-team lives at my-team.harnesser.co. Names are 1 to 30 characters of lowercase letters, digits, and hyphens, starting and ending with a letter or digit. The wizard checks availability as you type. A few names used by the platform itself (such as console, api, and www) are reserved.

5. Wait for provisioning

After you click Create harness, the console shows live progress through four steps:

1. Issuing model-gateway key. Your harness gets its own key for the built-in model gateway.
2. Creating your microVM. The dedicated VM, disk, and network rules are created.
3. Waiting for capacity. If the platform needs to add server capacity for you, this step can take a few minutes on first use.
4. Booting the harness. The harness software starts and comes online.

Provisioning usually finishes in under 3 minutes. You can leave the page; the build keeps going. If a step fails, the detail page shows the error with a Retry button. Retry rebuilds the harness from scratch and discards the disk from the failed attempt.

6. Open your workspace

When the status turns to running, open https://<name>.harnesser.co. New harnesses start private: only you, signed in to the console, can reach the URL. Each harness has its own first sign-in flow inside the workspace; the getting started page for Paperclip, QM, or Hermes walks through it.

Your disk persists between sessions. Close the browser, come back later, and your workspaces, files, and data are still there.

Deleting a harness

Deleting is permanent. It destroys the microVM, the disk with all workspaces and data, and frees the <name>.harnesser.co address. The console asks you to type the harness name to confirm. There is no undo and no backup, so export anything you need first.