AI Foundry Hosted Agent A2A stopped working HOSTED_AGENT_NOT_SUPPORTED

Phuong Le 0 Reputation points
2026-10-26T12:47:53.0366667+00:00

We have encountered an issue with incoming A2A (Agent-to-Agent) requests to Microsoft Foundry Hosted Agents that previously worked successfully.

Our Hosted Agent was successfully receiving A2A message/send requests through September 21, 2026, as confirmed by our application traces.

Since then, the same agent deployment and A2A endpoint have stopped accepting requests, without any code changes, redeployment, or configuration changes on our side.

Error

Both A2A protocol versions 0.3 and 1.0 return the same error:

{
  "jsonrpc": "2.0",
  "error": {
    "code": -32099,
    "message": "The requested A2A operation is not supported for this hosted-agent target.",
    "data": {
      "code": "HostedAgentNotSupported",
      "agentName": "agent-routing-service",
      "detail": "The requested A2A operation is not supported for hosted-agent target 'agent-routing-service'. Use a prompt agent as the A2A target."
    }
  }
}

For A2A v1.0, the equivalent error reason is HOSTED_AGENT_NOT_SUPPORTED.

Documentation discrepancy

Microsoft's Hosted Agents documentation describes A2A endpoints for Hosted Agents.

However, the Enable incoming A2A documentation identifies Prompt Agents as supported incoming A2A targets.

We have not found an official announcement explaining why incoming A2A requests that previously worked against Hosted Agents are now rejected.

Questions

Was incoming A2A support for Hosted Agents intentionally removed or restricted?

Is Microsoft planning to restore incoming A2A support for Hosted Agents?

We also found another report of the same behavior in this Microsoft Q&A discussion, where failures were reported beginning in late September.

Foundry Agent Service
Foundry Agent Service

A fully managed platform in Microsoft Foundry for hosting, scaling, and securing AI agents built with any supported framework or model

0 comments

1 answer

Sort by: Most helpful
Most helpful Newest Oldest
  1. Rukshan edirisinghe 2,445 Reputation points
    2026-10-26T13:42:28.15+00:00

    Hi @Phuong Le

    Your reading of the docs matches the error. The incoming A2A article (updated in September) says incoming A2A needs the responses protocol, and that prompt agents support it and can be exposed as A2A endpoints. It doesn't list hosted agents as incoming targets. The error's own detail line says the same thing: "Use a prompt agent as the A2A target." So this looks like the service now enforcing that rule, not something your deployment broke.

    To get callers working again without rewriting the hosted agent:

    1. Create a small prompt agent that acts as the A2A front door, with an agent card describing the same skills.
    2. Give that prompt agent a way to reach your hosted agent: for example, a tool that calls the hosted agent's own endpoint. It forwards the request and returns the answer.
    3. Enable A2A on the prompt agent (REST API or Python SDK, since the portal doesn't support it yet) and point your callers at its A2A endpoint. Use v1.0 with JSON-RPC for production, since 0.3 is preview.
    4. Grant your callers the Foundry Agent Consumer role on the project or that agent. A2A requires Entra ID auth, not keys.

    Also use Suggest a fix on the Hosted Agents page about the conflicting A2A description, so the docs get corrected.

    Do your A2A callers only send text, or do they also send files? Incoming A2A currently supports text only.

    If this gets your A2A traffic flowing again, please consider accepting it as the answer. If the prompt agent route is also rejected, let me know and I'll be happy to keep helping.

    Reference:

    Was this answer helpful?

    0 comments

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.