Skip to content

Harnesses we run

Different work wants a different harness. What does not change is the operating layer underneath, applied the same way to every one of them.

A continuous unbroken process running above layers of accumulated state.

OpenClaw

Available now

The agent that shows up for work

A long-running agent harness built around persistent employees. Hire one, give it tools and a personality, and it keeps working between conversations.

  • Persistent employees with durable memory across sessions
  • Personality and role configuration per employee
Identical units passing one at a time through a gate, carrying nothing between them.

Hermes

Available now

Fast, stateless, built to be called

A lean harness for agents that run on demand. Low cold-start, no session state to manage, and a clean surface for wiring into existing systems.

  • Per-request isolation with no shared state between runs
  • Low cold-start on managed infrastructure
A firm boundary drawn around a running core.

deepseek-harness

In preparation

Open weights, managed the same way

An open-weights agent harness for teams that need the model itself to run under their own terms. Not launched yet: join the list to shape it.

  • Open-weights models with no third-party inference hop
  • Region-pinned deployment for data residency

Not seeing yours?

The lineup grows with demand. Tell us what you run and we will tell you what it would take to run it for you.

Talk to an operator