Remote Agent (A2A) tools¶
Status: Current | Last reviewed: 2026-06-04
A Remote Agent tool lets one of your agents hand a task to another agent over the Agent-to-Agent (A2A) protocol. The remote agent publishes an agent card describing the skills it offers; you point TrueServe at it, choose which skills to enable, and your agent can delegate to those skills mid-conversation.
This works on any agent, regardless of its model — voice or chat, OpenAI, Anthropic, Google, or a realtime speech model. The remote agent's skills appear to your agent as ordinary tools it can call. This is how you build several specialised agents that collaborate on a complex problem: give one agent Remote Agent tools pointing at the others.
You can point a Remote Agent tool at a TrueServe agent you've exposed yourself — see Exposing an agent as A2A — so two of your own agents can talk to each other.
Configuration¶
| Field | Required | Notes |
|---|---|---|
| Agent URL | Yes | The remote agent's base endpoint, e.g. https://agents.example.com/support. |
| Authentication | No | Same auth options as other tools — see Tools overview. |
| Enabled skills | No | After TrueServe discovers the remote agent's skills from its agent card, pick the ones to expose. Leave empty to expose all discovered skills. |
When you save with a reachable Agent URL, TrueServe fetches the remote agent's card and caches its skill list so you can select among them. If the remote agent adds or changes skills later, re-open the tool to refresh.