Omnara ✓

Omnara is an open-source platform — positioned as the open alternative to Claude Managed Agents — that provides a single API for building agents that are durable, secure, and accessible to your team and customers.

Its premise is that a production agent is a service that has to be reachable, reliable, and secure, and Omnara delivers those guarantees so you deploy agents rather than infrastructure: you define the model, machines, tools, and permissions, and Omnara owns hosting, state, and durable execution.

Agents can run for minutes, hours, or days, survive crashes, and resume exactly where they left off, because every turn, tool call, and approval is durable state in a replayable event log rather than memory in a process.

Machines are live tools — managed-pool VMs agents create and destroy on demand, or your own connected devices and environments. Governance is built in: role-based access, scoped credentials, envelope-encrypted secrets (AES-256-GCM), platform-enforced allow/ask/deny tool policies, human approvals for risky actions, and an append-only, database-enforced audit history.

It's model-agnostic (OpenAI, Anthropic, OpenRouter, or self-hosted endpoints), supports MCP, versioned skills, checkpoints, and durable artifacts, and is fully open source and self-hostable so your data stays a database you can query, export, and own.

Backed by Y Combinator; desktop and mobile apps live at remote.omnara.com.