A2A Call

Tip

This document describes the detailed usage of a specific policy. If you are using the Apinizer policy structure for the first time or want to learn how policies work, read What is Policy? first.

Overview

What is its Purpose?

• Calls the A2A skill selected in advance by the administrator on every matching request.
• Does not leave skill selection to the LLM.
• Can continue without waiting for the result, or write the result to a variable.

Working Principle

1. The Agent Card is read from the agent endpoint.
2. Skills are discovered and targetSkillName is selected.
3. Gateway variables inside the Message Template are resolved.
4. The message is sent to the target skill.
5. FIRE_AND_FORGET does not wait; SYNC_TO_VARIABLE waits for the result and writes it to a variable.

Features and Capabilities

Basic Features

• Fixed skill: targetSkillName is required on the Fixed variant.
• Message template: Request body, header, and correlationId variables can be used.
• Wait mode: Continue without waiting, or write the result to an output variable.
• Fail-closed: The call is not allowed when discovery is empty or the target skill is not saved.

Advanced Features

• Private network: Allow Private-Network Egress must be enabled for a private-network target. Loopback, link-local, and metadata addresses are always forbidden.
• Timeouts: Call Timeout limits each HTTP send; Task Wait Timeout limits the whole task wait.
• Error: connectionConfigA2aRef or targetSkillName not configured means the agent or target skill is missing from the runtime configuration.

Usage Scenarios

Scenario	Situation	Solution (Policy Application)	Expected Behavior / Result
Fixed summary skill	The same skill must run on every chat request	summarize_report is selected; the message template forwards the user text from the request body	With fire-and-forget, the main request continues without waiting for the agent result

Configuring Policy Parameters

Agent Endpoint and Discovery

• Enter a base URL that can be resolved; do not append the Agent Card path.
• After discovery, the skill count must be greater than zero.
• Enable Allow Private-Network Egress for a private-network target.
• Loopback, link-local, and metadata addresses are always forbidden.

Target Skill and Message Template

• targetSkillName is required on the Fixed variant.
• Request body, header, and correlationId variables can be used in the Message Template.
• Confirm that the selected skill is still selected when the edit screen is opened again.

Wait Mode and Timeout

• FIRE_AND_FORGET: Does not wait for the result.
• SYNC_TO_VARIABLE: Waits for the result and writes it to the output variable.
• Call Timeout limits each HTTP send; Task Wait Timeout limits the whole task wait.

Fail-Closed Behavior

The call is not allowed when discovery is empty or the target skill is not saved.

Checklist

• Can the endpoint be resolved?
• Is private egress enabled when required?
• Did skill discovery succeed?
• Was the target skill selected and saved?
• Is the selection kept on the edit screen?
• Are wait mode and timeouts correct?
• Was the API Proxy redeployed?

Deleting the Policy

For deletion steps and what to do while the policy is in use, see Policy Management.

Exporting and Importing the Policy

For export and import steps, see Export/Import.

Connecting the Policy to an API

For how to attach this policy to APIs, see Policy Management.