Skip to main content
Each tool on an agent runs in one of three phases, and the phase decides when it runs and what data it can read. In the dashboard: Agents (listed under Assistants in the sidebar) → open an agent → Advanced Settings → Tools. On a multi-prompt agent, pre-call and post-call tools are under Global Settings and on-call tools are attached to each Agent node.

Before you start

An agent

The agent the tools run for.

Custom variables

Optional: per-call values tools can send, such as mobile_number.
For an API tool you also need the endpoint’s URL and any auth header it expects. For an Integration tool, connect the app on the sidebar Tools page first; accounts are connected once per workspace.

The three phases

The Tools page has a tab per phase. Each tab has a side panel listing what you can add. On pre-call and post-call, the panel also lists Frequently used integrations: calendar, sheets and Slack apps you have connected, with a warning icon on an expired account.

Use on-call tools in the prompt

An on-call tool runs only when the agent decides to call it, so the prompt must say when. Type @ in the prompt editor and pick the tool; it is inserted as @||tool_name||. Pre-call results go in as @((tool_name.field)).

Manage tools from an AI assistant

The Ringg MCP server (connect it) can list an agent’s tools, add or change API and function tools, test them and their jq, and remove them.

Verify it worked

Test the agent, then open the call in Logs and read the transcript.
The first message contains the pre-call values, and the transcript shows each on-call tool’s Tool Execution Message where it ran.

Troubleshooting

A pre-call tool returned a code listed in Block call on status codes. Test the tool with the same input, or remove the code.

Next steps

Call settings

Call duration, calling window, voicemail and interruptions.

Test the agent

Place a test call and check the tool calls in Logs.

Attach a number

Put the agent on a phone number.