Skip to content
oma
· OMA · 1 min read

Introduction to Open Managed Agents

Why we're building an open-source meta-harness for AI agents — and why BYOK matters.

intro byok

Welcome to Open Managed Agents (OMA) — an open-source platform for running AI agent fleets, designed as a self-hostable alternative to Claude Managed Agents.

What’s a meta-harness?

A harness is the loop that runs an agent: read events, build context, call the model, dispatch tools, persist state, recover from crashes. Most teams end up writing one. They’re roughly the same.

A meta-harness is the platform that runs harnesses — sessions, sandboxes, event log, memory, vaults, tools, integrations. The boring infrastructure your harness needs but doesn’t want to own.

OMA is a meta-harness. Write a harness. Deploy. The platform handles the rest.

Why BYOK

We don’t want to be in the business of marking up tokens. You already pay Anthropic / OpenAI / OpenRouter directly. Adding our own token margin on top would mean a worse rate for the same model, with no value added.

So we charge for what we actually run: the sandbox. Local runtime is free forever. Cloud sandbox is $0.005/min, billed in 1-minute increments.

What’s next

We’re publishing the cloud runtime + Console at oma.duyet.net now (early access) and the self-host story (Docker + SQLite or Postgres) shortly after. Star the repo to follow along.