Connect
New systems are new tools, not a new model.
As external systems connect — office suites, commerce platforms, payment providers, chat, whatever you already run — they arrive as more registry entries. The identity model, the permission model, the approval rules and the audit trail do not change, so your existing agents do not need rebuilding.
Your agents keep working
An agent written against "read customer" keeps working when a second customer system connects. It is holding a capability, not an integration.
External capabilities are not privileged
A tool from a large vendor gets no more trust than one you wrote. Same permissions, same approval behaviour, same audit record.
Rate limits and failures are the platform's problem
Retries, backoff and timeouts sit in the runtime rather than in each agent's logic. Otherwise every agent author re-implements them, differently.
Network reach stays restricted
Connecting an external system does not open general internet access. The agent can reach that system's declared endpoints and nothing else — see network policy under permissions.
Tools
The rest of this area.
Named capabilities with defined inputs, outputs and risk.
The tool registry
Every capability an agent could be given, in one list.
What a tool declares
Inputs, outputs, risk, permission, approval behaviour.
Risk levels
Four levels, and the policies that attach to them.
Your business as a toolbox
Connected systems become capabilities, not open access.
Your own tools
Expose your own service as a permissioned capability.
Give AI a job.
Build an agent, choose what it can access, define what it may do, and deploy it to work safely across your business.
No code required · Full SDKs and API · Stop any agent instantly · [email protected]