Agent Engine
The managed runtime. Any harness, any sandbox, your cloud.
Build, deploy, run, and govern managed agents on any harness, any model, any sandbox — no lock-in. Engine and Gateway are open source.
Most platforms bind you to one framework and one model. Orca is open at every layer — bring what you use, swap it with a config change, and never rewrite your agents.
Run the harness you already use — swap it with a config change, not a rewrite.
Route to any provider. Switching AI models never touches agent code.
Pluggable isolated execution for every tool call and code run.
Agents wake on real-time events, act on tools through a governed gateway, and write every step back as a replayable stream — accountable end to end.

Three open layers across the agent lifecycle: build and run, route and govern, measure and improve. Adopt one, or run them as one platform.
The managed runtime. Any harness, any sandbox, your cloud.
Vaulted credentials, budgets, and audit on every call.
Transcripts, live tailing, replay, and version lineage.
The Agent Engine and AI Gateway are open source under the Apache-2.0 license, released in Developer Preview. Run the whole stack in your own cloud, or let Orca Cloud run it for you — it's the same core either way. No open-core bait-and-switch, no proprietary agent format, no data leaving your workspace.