User/harsheetshah/toolbox call id forwarding

samples/csharp/hosted-agents/bring-your-own/invocations/HelloWorld/agent.yaml

@@ -3,7 +3,7 @@ kind: hosted
 name: hello-world-dotnet-invocations
 protocols:
   - protocol: invocations
-    version: 1.0.0
+    version: 2.0.0
 resources:
   cpu: "0.25"
   memory: 0.5Gi

samples/csharp/hosted-agents/bring-your-own/invocations/human-in-the-loop/agent.yaml

@@ -3,7 +3,7 @@ kind: hosted
 name: human-in-the-loop-dotnet-invocations
 protocols:
   - protocol: invocations
-    version: 1.0.0
+    version: 2.0.0
 resources:
   cpu: "0.25"
   memory: 0.5Gi

samples/csharp/hosted-agents/bring-your-own/invocations/notetaking-agent/agent.yaml

@@ -3,7 +3,7 @@ kind: hosted
 name: notetaking-agent-dotnet-invocations
 protocols:
   - protocol: invocations
-    version: 1.0.0
+    version: 2.0.0
 resources:
   cpu: "0.25"
   memory: 0.5Gi

samples/csharp/hosted-agents/bring-your-own/invocations_ws/HelloWorld/agent.yaml

@@ -3,7 +3,7 @@ kind: hosted
 name: hello-world-dotnet-invocations-ws
 protocols:
   - protocol: invocations_ws
-    version: 1.0.0
+    version: 2.0.0
 resources:
   cpu: "1"
   memory: 2Gi

samples/csharp/hosted-agents/bring-your-own/responses/HelloWorld/agent.yaml

@@ -3,7 +3,7 @@ kind: hosted
 name: hello-world-dotnet-responses
 protocols:
   - protocol: responses
-    version: 1.0.0
+    version: 2.0.0
 resources:
   cpu: "0.25"
   memory: 0.5Gi

samples/csharp/hosted-agents/bring-your-own/responses/background-agent/agent.yaml

@@ -3,7 +3,7 @@ kind: hosted
 name: background-agent-dotnet-responses
 protocols:
   - protocol: responses
-    version: 1.0.0
+    version: 2.0.0
 resources:
   cpu: "0.25"
   memory: 0.5Gi

samples/csharp/hosted-agents/bring-your-own/responses/browser-automation/agent.yaml

@@ -13,7 +13,7 @@ metadata:
         - .NET
 protocols:
     - protocol: responses
-      version: 1.0.0
+      version: 2.0.0
 resources:
     cpu: "0.25"
     memory: 0.5Gi

samples/csharp/hosted-agents/bring-your-own/responses/env-vars-agent/agent.yaml

@@ -17,7 +17,7 @@ metadata:
         - .NET
 protocols:
     - protocol: responses
-      version: 1.0.0
+      version: 2.0.0
 resources:
     cpu: "0.25"
     memory: 0.5Gi

samples/csharp/hosted-agents/bring-your-own/responses/notetaking-agent/agent.yaml

@@ -3,7 +3,7 @@ kind: hosted
 name: notetaking-agent-dotnet-responses
 protocols:
   - protocol: responses
-    version: 1.0.0
+    version: 2.0.0
 resources:
   cpu: "0.25"
   memory: 0.5Gi

samples/python/hosted-agents/bring-your-own/invocations/ag-ui/agent.yaml

@@ -3,7 +3,7 @@ kind: hosted
 name: ag-ui-invocations
 protocols:
   - protocol: invocations
-    version: 1.0.0
+    version: 2.0.0
 resources:
   cpu: "0.25"
   memory: 0.5Gi

samples/python/hosted-agents/bring-your-own/invocations/claude-agent-sdk/agent.yaml

@@ -13,7 +13,7 @@ metadata:
         - Bring Your Own
 protocols:
     - protocol: invocations
-      version: 1.0.0
+      version: 2.0.0
 resources:
     cpu: "0.25"
     memory: 0.5Gi

samples/python/hosted-agents/bring-your-own/invocations/diagnostic-agent/agent.yaml

@@ -3,7 +3,7 @@ kind: hosted
 name: diagnostic-agent-python-invocations
 protocols:
   - protocol: invocations
-    version: 1.0.0
+    version: 2.0.0
 resources:
   cpu: "0.25"
   memory: 0.5Gi

samples/python/hosted-agents/bring-your-own/invocations/event-grid-trigger/agent.yaml

@@ -5,7 +5,7 @@ description: |
   BYO Invocations agent that Event Grid POSTs to directly using the system topic's system-assigned managed identity with audience https://ai.azure.com. Receives Storage BlobCreated events, downloads the blob with the per-agent Microsoft Entra identity, and writes a summary back to a sibling container.
 protocols:
   - protocol: invocations
-    version: 1.0.0
+    version: 2.0.0
 resources:
   cpu: "0.25"
   memory: 0.5Gi

samples/python/hosted-agents/bring-your-own/invocations/github-copilot/agent.yaml

@@ -3,7 +3,7 @@ kind: hosted
 name: github-copilot-invocations
 protocols:
   - protocol: invocations
-    version: 1.0.0
+    version: 2.0.0
 resources:
   cpu: '0.25'
   memory: 0.5Gi

samples/python/hosted-agents/bring-your-own/invocations/hello-world/agent.yaml

@@ -3,7 +3,7 @@ kind: hosted
 name: hello-world-python-invocations
 protocols:
   - protocol: invocations
-    version: 1.0.0
+    version: 2.0.0
 resources:
   cpu: "0.25"
   memory: 0.5Gi

samples/python/hosted-agents/bring-your-own/invocations/human-in-the-loop/agent.yaml

@@ -3,7 +3,7 @@ kind: hosted
 name: human-in-the-loop-invocations
 protocols:
   - protocol: invocations
-    version: 1.0.0
+    version: 2.0.0
 resources:
   cpu: "0.25"
   memory: 0.5Gi

samples/python/hosted-agents/bring-your-own/invocations/langgraph-chat/agent.yaml

@@ -14,7 +14,7 @@ metadata:
         - Python
 protocols:
     - protocol: invocations
-      version: 1.0.0
+      version: 2.0.0
 resources:
     cpu: "0.25"
     memory: 0.5Gi

samples/python/hosted-agents/bring-your-own/invocations/notetaking-agent/agent.yaml

@@ -2,7 +2,7 @@ kind: hosted
 name: notetaking-agent-invocations-python
 protocols:
   - protocol: invocations
-    version: 1.0.0
+    version: 2.0.0
 resources:
   cpu: "0.25"
   memory: 0.5Gi

samples/python/hosted-agents/bring-your-own/invocations/toolbox/agent.yaml

@@ -3,7 +3,7 @@ kind: hosted
 name: toolbox-python-invocations
 protocols:
   - protocol: invocations
-    version: 1.0.0
+    version: 2.0.0
 resources:
   cpu: "0.25"
   memory: 0.5Gi

samples/python/hosted-agents/bring-your-own/invocations/toolbox/main.py

@@ -118,6 +118,12 @@
 _TOOLBOX_FEATURES = os.getenv(
     "FOUNDRY_AGENT_TOOLBOX_FEATURES", "Toolboxes=V1Preview")
 
+# Platform-injected per-request call identifier (container protocol v2.0.0).
+# Extracted from the inbound invocation request and forwarded verbatim on every
+# egress call to the Foundry toolbox MCP proxy so the platform can correlate the
+# downstream tool calls with the originating invocation. Never parse this value.
+_FOUNDRY_CALL_ID_HEADER = "x-agent-foundry-call-id"
+
 _credential = DefaultAzureCredential()
 _project_client = AIProjectClient(endpoint=_endpoint, credential=_credential)
 _responses_client = _project_client.get_openai_client().responses
@@ -142,13 +148,16 @@ def __init__(self, endpoint: str, token_provider):
         self._session_id: str | None = None
         self._req_id = 0
 
-    def _headers(self) -> dict:
+    def _headers(self, call_id: str | None = None) -> dict:
         h = {
             "Content-Type": "application/json",
             "Authorization": f"Bearer {self._get_token()}",
         }
         if _TOOLBOX_FEATURES:
             h["Foundry-Features"] = _TOOLBOX_FEATURES
+        # Forward the per-request call ID extracted from the inbound invocation.
+        if call_id:
+            h[_FOUNDRY_CALL_ID_HEADER] = call_id
         if self._session_id:
             h["mcp-session-id"] = self._session_id
         return h
@@ -157,12 +166,12 @@ def _next_id(self) -> int:
         self._req_id += 1
         return self._req_id
 
-    def initialize(self) -> str:
+    def initialize(self, call_id: str | None = None) -> str:
         """Send MCP initialize + initialized notification."""
         with httpx.Client(timeout=60) as client:
             resp = client.post(
                 self.endpoint,
-                headers=self._headers(),
+                headers=self._headers(call_id),
                 json={
                     "jsonrpc": "2.0",
                     "id": self._next_id(),
@@ -181,29 +190,29 @@ def initialize(self) -> str:
             # Send initialized notification
             client.post(
                 self.endpoint,
-                headers=self._headers(),
+                headers=self._headers(call_id),
                 json={"jsonrpc": "2.0", "method": "notifications/initialized"},
             )
             return data.get("result", {}).get("serverInfo", {}).get("name", "unknown")
 
-    def list_tools(self) -> list[dict]:
+    def list_tools(self, call_id: str | None = None) -> list[dict]:
         """Call tools/list and return tool definitions."""
         with httpx.Client(timeout=60) as client:
             resp = client.post(
                 self.endpoint,
-                headers=self._headers(),
+                headers=self._headers(call_id),
                 json={"jsonrpc": "2.0", "id": self._next_id(
                 ), "method": "tools/list", "params": {}},
             )
             resp.raise_for_status()
             return resp.json().get("result", {}).get("tools", [])
 
-    def call_tool(self, name: str, arguments: dict) -> str:
+    def call_tool(self, name: str, arguments: dict, call_id: str | None = None) -> str:
         """Call a tool and return the text result."""
         with httpx.Client(timeout=120) as client:
             resp = client.post(
                 self.endpoint,
-                headers=self._headers(),
+                headers=self._headers(call_id),
                 json={
                     "jsonrpc": "2.0",
                     "id": self._next_id(),
@@ -235,13 +244,13 @@ def call_tool(self, name: str, arguments: dict) -> str:
 _tools_initialized = False
 
 
-def _ensure_tools():
+def _ensure_tools(call_id: str | None = None):
     global _mcp_client, _tool_definitions, _tools_initialized
     if _tools_initialized:
         return
     _mcp_client = _McpToolboxClient(TOOLBOX_ENDPOINT, _token_provider)
-    server_name = _mcp_client.initialize()
-    mcp_tools = _mcp_client.list_tools()
+    server_name = _mcp_client.initialize(call_id)
+    mcp_tools = _mcp_client.list_tools(call_id)
     logger.info("Toolbox '%s' connected: %d tool(s) discovered",
                 server_name, len(mcp_tools))
     for t in mcp_tools:
@@ -265,9 +274,9 @@ def _ensure_tools():
 _MAX_TOOL_ROUNDS = 10
 
 
-def _call_model(input_items: list[dict]) -> object:
+def _call_model(input_items: list[dict], call_id: str | None = None) -> object:
     """Call the model with tool definitions and return the response."""
-    _ensure_tools()
+    _ensure_tools(call_id)
     return _responses_client.create(
         model=_model,
         instructions=_SYSTEM_PROMPT,
@@ -277,15 +286,15 @@ def _call_model(input_items: list[dict]) -> object:
     )
 
 
-def _run_agent_loop(input_items: list[dict]) -> str:
+def _run_agent_loop(input_items: list[dict], call_id: str | None = None) -> str:
     """Execute the agentic tool-calling loop synchronously.
 
     Calls the model, checks for tool calls, executes them, feeds results
     back, and repeats until the model produces a text response or we hit
     the max rounds limit.
     """
     for _ in range(_MAX_TOOL_ROUNDS):
-        response = _call_model(input_items)
+        response = _call_model(input_items, call_id)
 
         # Check if the model wants to call tools
         tool_calls = [
@@ -302,7 +311,7 @@ def _run_agent_loop(input_items: list[dict]) -> str:
             try:
                 arguments = json.loads(tc.arguments) if isinstance(
                     tc.arguments, str) else tc.arguments
-                result_text = _mcp_client.call_tool(tc.name, arguments)
+                result_text = _mcp_client.call_tool(tc.name, arguments, call_id)
                 logger.info("Tool '%s' returned %d chars",
                             tc.name, len(result_text))
             except Exception as e:
@@ -326,10 +335,10 @@ def _run_agent_loop(input_items: list[dict]) -> str:
     return "(Reached maximum tool call rounds)"
 
 
-async def _stream_agent_reply(input_items: list[dict]):
+async def _stream_agent_reply(input_items: list[dict], call_id: str | None = None):
     """Run the agent loop in a thread and yield the result as SSE events."""
     loop = asyncio.get_running_loop()
-    result = await loop.run_in_executor(None, _run_agent_loop, input_items)
+    result = await loop.run_in_executor(None, _run_agent_loop, input_items, call_id)
     yield f"data: {json.dumps({'type': 'token', 'content': result})}\n\n"
 
 
@@ -374,8 +383,12 @@ async def handle_invoke(request: Request):
     session_id = request.state.session_id
     invocation_id = request.state.invocation_id
 
-    logger.info("Processing invocation %s (session %s)",
-                invocation_id, session_id)
+    # Extract the per-request call ID (container protocol v2.0.0) directly from
+    # the inbound request headers so it can be forwarded on toolbox egress calls.
+    call_id = request.headers.get(_FOUNDRY_CALL_ID_HEADER)
+
+    logger.info("Processing invocation %s (session %s, call_id %s)",
+                invocation_id, session_id, call_id)
 
     # Retrieve or create conversation history for this session.
     history = _sessions.setdefault(session_id, [])
@@ -385,7 +398,7 @@ async def handle_invoke(request: Request):
     async def event_generator():
         full_reply = ""
         try:
-            async for delta in _stream_agent_reply(input_items):
+            async for delta in _stream_agent_reply(input_items, call_id):
                 # Parse the SSE data to extract the content for history
                 try:
                     event_data = json.loads(delta.split(

samples/python/hosted-agents/bring-your-own/invocations_ws/duplex-live-agent/agent.yaml

@@ -3,7 +3,7 @@ kind: hosted
 name: duplex-live-agent
 protocols:
   - protocol: invocations_ws
-    version: 1.0.0
+    version: 2.0.0
 resources:
   cpu: "1"
   memory: 1Gi

samples/python/hosted-agents/bring-your-own/invocations_ws/hello-world/agent.yaml

@@ -3,7 +3,7 @@ kind: hosted
 name: hello-world
 protocols:
   - protocol: invocations_ws
-    version: 1.0.0
+    version: 2.0.0
 resources:
   cpu: "1"
   memory: 2Gi

samples/python/hosted-agents/bring-your-own/invocations_ws/livekit-server/agent.yaml

@@ -3,7 +3,7 @@ kind: hosted
 name: livekit-server
 protocols:
   - protocol: invocations_ws
-    version: 1.0.0
+    version: 2.0.0
 resources:
   cpu: "2"
   memory: 4Gi

samples/python/hosted-agents/bring-your-own/invocations_ws/pipecat-webrtc/agent.yaml

@@ -3,7 +3,7 @@ kind: hosted
 name: pipecat-ws-webrtc
 protocols:
   - protocol: invocations_ws
-    version: 1.0.0
+    version: 2.0.0
 resources:
   cpu: "2"
   memory: 4Gi

samples/python/hosted-agents/bring-your-own/invocations_ws/pipecat-ws-server/agent.yaml

@@ -3,7 +3,7 @@ kind: hosted
 name: pipecat-ws-server
 protocols:
   - protocol: invocations_ws
-    version: 1.0.0
+    version: 2.0.0
 resources:
   cpu: "2"
   memory: 4Gi

samples/python/hosted-agents/bring-your-own/responses/background-agent/agent.yaml

@@ -3,7 +3,7 @@ kind: hosted
 name: background-agent-responses
 protocols:
   - protocol: responses
-    version: 1.0.0
+    version: 2.0.0
 resources:
   cpu: "0.25"
   memory: 0.5Gi