Connect
Give agents controlled ways to work.
A tool is one capability with a name, a defined input, a defined output and a risk level — "read an invoice", "send an approved reminder". An agent can call the tools it has been granted and has no other route into your systems. It does not hold your credentials, and it cannot reach a capability nobody gave it.
Capabilities, not credentials
The agent never sees an API key, a connection string or a password. It calls a capability by name and the platform performs the call on its behalf, with its identity attached.
Narrow on purpose
"Read an invoice" is a tool. "Query the finance database" is not, and never will be, because the second cannot be reviewed — its risk depends entirely on what the agent decides to ask for.
Every tool publishes its risk
Inputs, outputs, risk level, the permission it requires and whether it needs approval. A capability that has not declared these cannot be added to an agent.
In this area
6 pages under Tools.
Each one is its own surface in the product, and each one is here because it does something the others do not.
The tool registry
Every capability an agent could be given, in one list.
ReadWhat a tool declares
Inputs, outputs, risk, permission, approval behaviour.
ReadRisk levels
Four levels, and the policies that attach to them.
ReadYour business as a toolbox
Connected systems become capabilities, not open access.
ReadYour own tools
Expose your own service as a permissioned capability.
ReadAdding external systems
The same architecture, more tools. No redesign.
ReadThe rest of the platform
This is one part of one loop.
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]