Giving agents tools
Let an agent act — call integrations, run workflows, and hand off to other agents.
A tool is a capability you grant an agent so it can do things, not just respond. With tools, an agent can query a database, call an API, run a whole workflow, or hand part of the job to another agent. The agent decides when to use each tool while it works; you decide which tools it's allowed to use.
The three kinds of tool
Integration actions
Specific operations on a connected service — query a table, call an endpoint, send a message.
Workflows
An entire workflow, callable as a single tool. The agent passes inputs and gets the workflow's output back.
Other agents
Another agent, so one agent can delegate a sub-task to a specialist.
Adding a tool
In the builder's Tools panel, click Add tool. A picker opens listing everything available in the project, grouped by kind:
Choose what to add
Search and pick an integration, a workflow, or another agent.
For an integration, choose its actions
Select which of the integration's actions the agent may call. You can grant one, a few, or all of them.
Make sure it has a connection
An integration tool needs a connection to run through. If one isn't set, the picker flags it — pick a connection to resolve the warning.
The agent's Tools panel then lists what it can use. Remove a tool any time to take the capability away.
Tools that are other agents or workflows
When you add a workflow or another agent as a tool, the agent can invoke it by name, pass it input, and use what it returns — all within a single run. This is how you compose bigger behavior out of smaller, focused pieces.
Tools must be deployed
A workflow or agent has to be deployed before another agent can use it as a tool. Drafts aren't callable — deploy the piece you want to reuse first, then add it as a tool.
How tools show up in a run
Every tool call the agent makes is recorded in the run's trace: which tool, the inputs it was called with, and what it returned. When a tool is itself a workflow or agent, you can expand its trace to see that sub-run's own steps inline. This makes it easy to follow exactly how the agent reached its result.