Python: Emit AG-UI events for MCP tool calls, results, and text reasoning by LEDazzio01 · Pull Request #4233 · microsoft/agent-framework

LEDazzio01 · 2026-02-24T23:05:28Z

Summary

Fixes #4213 — _emit_content() in the AG-UI layer only handled text, function_call, function_result, function_approval_request, and usage content types. Foundry MCP content types (mcp_server_tool_call, mcp_server_tool_result) and text_reasoning fell through unhandled, producing no SSE events for AG-UI consumers.

Changes

`_run_common.py`

Added three new handler functions and wired them into _emit_content():

Content Type	Handler	AG-UI Events
`mcp_server_tool_call`	`_emit_mcp_tool_call`	`TOOL_CALL_START` + `TOOL_CALL_ARGS`
`mcp_server_tool_result`	`_emit_mcp_tool_result`	`TOOL_CALL_END` + `TOOL_CALL_RESULT`
`text_reasoning`	`_emit_text_reasoning`	`CUSTOM` (name=`"text_reasoning"`)

Design decisions:

MCP tool calls are emitted as complete start→args sequences (not streamed) because Foundry sends them as fully-formed content items
server_name/tool_name format for display name when server_name is available, for disambiguation
text_reasoning uses CustomEvent since AG-UI protocol has no dedicated reasoning event type — consistent with how _emit_usage works
MCP tool calls/results are tracked in FlowState (pending_tool_calls, tool_calls_by_id, tool_results, tool_calls_ended) for proper MESSAGES_SNAPSHOT inclusion

`test_run.py`

Added 15 test cases across 4 new test classes:

TestEmitMcpToolCall — start/args events, flow tracking, server name prefix, missing args, generated IDs
TestEmitMcpToolResult — end/result events, flow tracking, missing call_id, JSON serialization
TestEmitTextReasoning — custom event, protected_data fallback, empty handling, optional id
TestEmitContentMcpRouting — verifies _emit_content dispatch for all three new content types

…t reasoning Fixes microsoft#4213 — mcp_server_tool_call, mcp_server_tool_result, and text_reasoning content types now produce AG-UI SSE events instead of being silently dropped by _emit_content(). - _emit_mcp_tool_call: maps to ToolCallStart + ToolCallArgs events - _emit_mcp_tool_result: maps to ToolCallEnd + ToolCallResult events - _emit_text_reasoning: maps to CustomEvent(name="text_reasoning")

Tests cover: - _emit_mcp_tool_call: start/args events, flow tracking, server name prefix - _emit_mcp_tool_result: end/result events, flow tracking, serialization - _emit_text_reasoning: custom event, protected_data fallback, empty handling - _emit_content routing: verifies dispatch for all three new content types

Copilot

Pull request overview

Adds missing AG-UI SSE emission for Foundry-specific content types so MCP tool activity and text_reasoning are visible to AG-UI consumers (fixes #4213).

Changes:

Add _emit_mcp_tool_call, _emit_mcp_tool_result, and _emit_text_reasoning, and route them from _emit_content().
Track MCP tool calls/results in FlowState for MESSAGES_SNAPSHOT inclusion.
Add new unit tests covering MCP call/result emission and text_reasoning routing.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 3 comments.

File	Description
python/packages/ag-ui/agent_framework_ag_ui/_run_common.py	Implements new emitters for MCP tool call/result and `text_reasoning`, and wires dispatch in `_emit_content()`.
python/packages/ag-ui/tests/ag_ui/test_run.py	Adds test coverage for new emitters and routing behavior.

python/packages/ag-ui/agent_framework_ag_ui/_run_common.py

Copilot · 2026-02-24T23:10:48Z

python/packages/ag-ui/agent_framework_ag_ui/_run_common.py

+    events.append(ToolCallEndEvent(tool_call_id=content.call_id))
+    flow.tool_calls_ended.add(content.call_id)
+
+    raw_output = content.output if content.output is not None else ""
+    result_content = raw_output if isinstance(raw_output, str) else json.dumps(make_json_safe(raw_output))
+    message_id = generate_event_id()
+    events.append(
+        ToolCallResultEvent(
+            message_id=message_id,
+            tool_call_id=content.call_id,
+            content=result_content,
+            role="tool",
+        )
+    )
+
+    flow.tool_results.append(
+        {
+            "id": message_id,
+            "role": "tool",
+            "toolCallId": content.call_id,
+            "content": result_content,
+        }
+    )
+
+    return events


_emit_mcp_tool_result doesn’t perform the same FlowState cleanup as _emit_tool_result (e.g., applying predictive_handler updates, resetting tool_call_id/tool_call_name, and closing/resetting an open text message_id/accumulated_text). This can leave a text message “open” across MCP tool results and can make MessagesSnapshotEvent IDs/content inconsistent compared to function_result handling. Consider mirroring _emit_tool_result’s cleanup logic (or refactoring to share it) so MCP results behave the same as standard tool results.

1. _emit_text_reasoning: Don't fall back to protected_data as display text — expose it under a separate key so consumers can decide how to render it (avoids leaking provider-specific metadata as text). 2. _emit_mcp_tool_call: Fix docstring to accurately describe behavior (start + args only; end handled by _emit_mcp_tool_result). 3. _emit_mcp_tool_result: Mirror _emit_tool_result cleanup logic — reset tool_call_id/tool_call_name, close open text messages, reset accumulated_text so MCP results behave consistently with standard tool results.

- TestEmitTextReasoning: test protected_data as separate key (not fallback text), test protected_data-only content still emits event - TestEmitMcpToolResult: add tests for FlowState cleanup (message closing, tool_call_id/name reset, accumulated_text reset)

LEDazzio01

Thanks for the thorough review! All 3 comments have been addressed in the latest commits:

protected_data: No longer used as fallback display text — exposed under a separate "protected_data" key so consumers can decide how to render it
Docstring: Updated to accurately describe start/args-only behavior
FlowState cleanup: _emit_mcp_tool_result now mirrors _emit_tool_result — resets tool_call_id/tool_call_name, closes open text messages, resets accumulated_text

Updated tests added to cover the new cleanup behavior and protected_data separation.

moonbox3 · 2026-02-25T09:51:45Z

python/packages/ag-ui/agent_framework_ag_ui/_run_common.py

+    execution visible to AG-UI consumers. Completion/end events are handled
+    separately by ``_emit_mcp_tool_result``.
+
+    Fixes #4213.


Please remove Fixes #4213.

moonbox3 · 2026-02-25T09:56:28Z

python/packages/ag-ui/agent_framework_ag_ui/_run_common.py

+    events: list[BaseEvent] = []
+
+    if not content.call_id:
+        return events


in the event that call_id is missing, we simply return. Should we log anything?

moonbox3 · 2026-02-25T09:57:19Z

python/packages/ag-ui/agent_framework_ag_ui/_run_common.py

+    return events
+
+
+def _emit_text_reasoning(content: Content, flow: FlowState) -> list[BaseEvent]:


flow is never used in this method?

markwallace-microsoft · 2026-02-25T10:01:53Z

Python Test Coverage Report •

File	Stmts	Miss	Cover	Missing
packages/ag-ui/agent_framework_ag_ui
_run_common.py	248	29	88%	46, 51–52, 54, 61, 65, 69–70, 72, 96, 173, 221–222, 236, 263–265, 290–291, 297–302, 522, 524, 533–534
TOTAL	21789	2760	87%

Python Unit Test Overview

Tests	Skipped	Failures	Errors	Time
4549	246 💤	0 ❌	0 🔥	1m 16s ⏱️

1. Remove `Fixes microsoft#4213` from all emitter docstrings 2. Add logger.warning when call_id is missing in _emit_mcp_tool_result 3. Remove unused `flow` param from _emit_text_reasoning

…ve (Fixes microsoft#4213) comment

LEDazzio01 · 2026-02-25T13:28:25Z

Thanks for the review @moonbox3! All 3 comments have been addressed in the latest commits:

Fixes #4213 — Removed from all three emitter docstrings (_emit_mcp_tool_call, _emit_mcp_tool_result, _emit_text_reasoning)
Missing call_id logging — Added logger.warning("MCP tool result content missing call_id, skipping") in _emit_mcp_tool_result when call_id is absent
Unused flow param — Removed flow from _emit_text_reasoning signature (it mirrors _emit_usage which also doesn't need flow state). Updated the _emit_content dispatcher and all test call sites accordingly.

LEDazzio01 added 2 commits February 24, 2026 18:02

Copilot AI review requested due to automatic review settings February 24, 2026 23:05

markwallace-microsoft added the python label Feb 24, 2026

Copilot started reviewing on behalf of LEDazzio01 February 24, 2026 23:06 View session

LEDazzio01 mentioned this pull request Feb 24, 2026

Python: [Bug]: MCP Tool Calls + Reasoning Not Emitted as AG-UI Events (Foundry) #4213

Open

Copilot AI reviewed Feb 24, 2026

View reviewed changes

LEDazzio01 added 2 commits February 24, 2026 18:14

LEDazzio01 commented Feb 24, 2026

View reviewed changes

moonbox3 reviewed Feb 25, 2026

View reviewed changes

LEDazzio01 added 2 commits February 25, 2026 08:25

Address review feedback from @moonbox3

c76f966

1. Remove `Fixes microsoft#4213` from all emitter docstrings 2. Add logger.warning when call_id is missing in _emit_mcp_tool_result 3. Remove unused `flow` param from _emit_text_reasoning

Update tests: remove flow param from _emit_text_reasoning calls, remo…

8122740

…ve (Fixes microsoft#4213) comment

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Comments

Python: Emit AG-UI events for MCP tool calls, results, and text reasoning#4233

Python: Emit AG-UI events for MCP tool calls, results, and text reasoning#4233
LEDazzio01 wants to merge 6 commits intomicrosoft:mainfrom
LEDazzio01:fix/4213-mcp-tool-call-events

LEDazzio01 commented Feb 24, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

Copilot AI Feb 24, 2026

Uh oh!

LEDazzio01 left a comment

Uh oh!

moonbox3 Feb 25, 2026

Uh oh!

moonbox3 Feb 25, 2026

Uh oh!

moonbox3 Feb 25, 2026

Uh oh!

markwallace-microsoft commented Feb 25, 2026

Uh oh!

LEDazzio01 commented Feb 25, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

		return events


		def _emit_text_reasoning(content: Content, flow: FlowState) -> list[BaseEvent]:

Comments

Conversation

LEDazzio01 commented Feb 24, 2026

Summary

Changes

_run_common.py

test_run.py

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Copilot AI Feb 24, 2026

Choose a reason for hiding this comment

Uh oh!

LEDazzio01 left a comment

Choose a reason for hiding this comment

Uh oh!

moonbox3 Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

moonbox3 Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

moonbox3 Feb 25, 2026

Choose a reason for hiding this comment

Uh oh!

markwallace-microsoft commented Feb 25, 2026

Python Unit Test Overview

Uh oh!

LEDazzio01 commented Feb 25, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

`_run_common.py`

`test_run.py`