The MCP tool connects your agent to a Model Context Protocol server. Whatever tools that server offers, the agent can use during the call: look up an order, create a record, run an action in your own systems.

​
Add it

Tools are added in Advanced; in Simple, the Tools step’s Open in Advanced takes you there.
  1. Open Components on the right rail and click MCP under Tools.
  2. Connect its output dot to the Persona (every Skill can use it) or to one Skill.
  3. Select the node and fill in the settings.

​
Settings

Voice.GradientFlow does not need a key of its own for MCP. The token you enter here goes to the server you named.

​
How the agent uses it

When the agent decides a tool is needed, it calls the server with the details it collected and reads the result back into the conversation. The tool shows a spinner on the canvas while it runs, and the transcript records the call. If the server has not answered within the node’s Time limit (seconds) (10 by default, under More settings), the agent stops waiting and tells the caller it could not finish. It does not say the action was done or try it again by itself, since the server may still have carried it out. To fill a long wait, set While running, say, for example “One moment while I check that.”; the agent says it once if the tool is still working after a second and a half.

​
Several servers

Add one MCP node per server. Connect them all to the Persona, or spread them across Skills so each phase only sees the tools it needs.

​
Tips

  • Tick only the tools the agent should have. Fewer tools means faster, more accurate decisions.
  • Give each Skill’s Instruction a sentence about when to use the tool: “Use the order lookup when the caller gives an order number.”