An OpenAgent is an agent built on an open stack: open tools it can discover and call, authenticated access it holds on your behalf, inspectable behavior, and no lock-in to a single model or vendor. Openness here is not the opposite of secure — it is the precondition for control and trust.

Open vs closed

A closed agent is a black box you rent: you cannot see its tools, move your context out, or reason about what it did. An OpenAgent is the inverse. Its tools are documented and callable. Its actions leave provenance. Its context is portable. You can swap the model underneath without losing the work above.

Why open earns trust

The more an agent touches — your memory, your customers, your money — the more you need to see and govern what it does. Open, authenticated tools make an agent's reach explicit: reads are free, writes are expected, and destructive or sharing actions require your consent. Trust comes from legibility, not from promises.

You can't outsource your intelligence

Every company is built on some specialized intelligence — its memory, its processes, its relationships. That is not something to hand to a black box you don't control. An open stack lets you keep improving your own intelligence in place, with agents you can inspect and direct.

The Cloud is open and authenticated

The Cloud's entire workspace is exposed as open tools over a standard interface, callable by any agent a user authorizes — and by humans working alongside them. Even onboarding is open: an agent can invite a team into a shared workspace through provisioned accounts. An Ora is an OpenAgent with governance built in.

Ora · Opus 4.8 · Cowork