GeezerVison is showing github.com — [REQUEST] AG-UI support · Issue #5969 · simstudioai/sim
[REQUEST] AG-UI support #5969DescriptionAdd support for the AG-UI (Agent-User Interaction) Protocol to enable standardized real-time communication between AI agents and frontend clients.
The implementation should expose agent execution events (e.g., messages, tool calls, reasoning/status updates, errors, completion, and streamed responses) using the AG-UI protocol, allowing compatible clients such as CopilotKit and other AG-UI consumers to interact with the backend without custom integrations.
Scope
- Implement AG-UI protocol support for agent execution.
- Stream agent events (text, tool calls, status, errors, completion, etc.).
- Map internal workflow/agent events to AG-UI event types.
- Support incremental/streaming responses.
- Ensure protocol compliance with the AG-UI specification.
- Provide an extensible architecture for future event types and protocol updates.
Acceptance Criteria
- AG-UI compatible clients can communicate with the backend.
- Agent responses are streamed using AG-UI events.
- Tool invocation lifecycle is exposed through the protocol.
- Errors and cancellation events are properly propagated.
- Protocol implementation is documented and covered by integration tests.
Benefits
- Standardized frontend integration.
- Interoperability with existing AG-UI ecosystem tools.
- Easier maintenance by decoupling internal agent logic from client-specific protocols.
- Foundation for future multi-agent and workflow integrations.
Activitywav-piyush commented on Sep 18, 2026 @ebe-wolf I dug into the cobebase, what i think is we should hold off on this AG-UI implementation for now.
I have some points
- we already have an internal streaming protocol system that covers streaming responses like tool lifecycles, errors, cancellation, checkpoints/resume, sub-agents, etc.
- there isn't a clear demand signal, yet i think If we had a concrete consumer like CopilotKit, that would be a strong reason to implement it. Without one, we'd mainly be adding compatibility for a potential future use case.
- there is an opportunity cost. Implementing this requires event mapping, streaming/tool handling, protocol compliance, tests, and documentation adding another API contract to maintain.
I think AG-UI is worth keeping on the roadmap rather than rejecting it outright.
@waleedlatif1 if you think we should add the AG-UI protocol.
Please let me know , i'm happy to implement it :)
MetadataMetadataDevelopmentNo branches or pull requestsIssue actions
|