@@ -14,12 +14,14 @@ This directory contains samples that demonstrate how to use [LangGraph](https://
 | 6   | [Files](responses/06-files/)                                          | A LangGraph agent with local filesystem tools and a Foundry-Toolbox `code_interpreter`, demonstrating session-uploaded file handling.                                                    |
 | 7   | [Human-in-the-Loop](responses/07-human-in-the-loop/)                  | A LangGraph `StateGraph` that drafts a proposal and pauses for human review via `langgraph.types.interrupt`, serialized as `mcp_approval_request` + `function_call` output items.        |
 | 8   | [Observability](responses/08-observability/)                          | A LangGraph agent with GenAI OpenTelemetry tracing enabled via `enable_auto_tracing()`, emitting spans, metrics, and logs to Application Insights.                                       |
+| 10  | [Run](responses/10-run/)                                              | A minimal configuration-driven LangGraph agent hosted over the Responses protocol with the `langchain_azure_ai.agents.hosting.run` entrypoint.                                         |
 
 ### Invocations API
 
 | #   | Sample                                          | Description                                                                                                                                          |
 | --- | ----------------------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------- |
 | 1   | [Chat](invocations/01-langgraph-chat/)          | A minimal LangGraph agent with two local tools, demonstrating session state via `agent_session_id` (URL param / `x-agent-session-id` response header) backed by a LangGraph checkpointer. |
+| 3   | [Run](invocations/03-run/)                      | A minimal configuration-driven LangGraph agent hosted over the Invocations protocol with the `langchain_azure_ai.agents.hosting.run` entrypoint and in-process `MemorySaver` state. |
 
 ### Agent-to-Agent (A2A)
 
 
@@ -0,0 +1,35 @@
+# Coding Agent Instructions
+
+This project is a **Microsoft Foundry hosted agent** built with LangGraph and the Invocations protocol. It runs in [Foundry Agent Service](https://learn.microsoft.com/en-us/azure/foundry/agents/concepts/hosted-agents) and uses the configuration-driven `langchain_azure_ai.agents.hosting.run` entrypoint.
+
+## Key files
+
+- `azure.yaml` - Foundry hosted-agent manifest
+- `src/langgraph-run-invocations/main.py` - exported LangGraph graph used by the hosting entrypoint
+- `src/langgraph-run-invocations/langgraph.json` - graph registry for the run command
+- `src/langgraph-run-invocations/requirements.txt` - locked dependencies
+
+## Development workflow
+
+```bash
+azd ai agent run --no-client
+azd ai agent invoke --local '{"message": "your message"}'
+azd deploy
+azd ai agent invoke '{"message": "your message"}'
+```
+
+## Microsoft Foundry Skill
+
+This project was built with the microsoft-foundry skill. Before working on or answering questions about Foundry agents, read that skill first.
+
+Install the skill with:
+
+```bash
+npx skills add https://github.com/microsoft/azure-skills --skill microsoft-foundry
+```
+
+## References
+
+- [Hosted agents overview](https://learn.microsoft.com/en-us/azure/foundry/agents/concepts/hosted-agents)
+- [LangGraph](https://langchain-ai.github.io/langgraph/)
+- [Microsoft Foundry Skill](https://learn.microsoft.com/en-us/azure/foundry/how-to/develop/use-microsoft-foundry-skill)
@@ -0,0 +1,5 @@
+# CLAUDE.md
+
+This project uses [AGENTS.md](./AGENTS.md) as the single source of truth for coding agent instructions. The import below loads it into Claude Code's context.
+
+@AGENTS.md
@@ -0,0 +1,153 @@
+# What this sample demonstrates
+
+A [LangGraph](https://langchain-ai.github.io/langgraph/) agent hosted using the **Invocations protocol** with session management. It uses the configuration-driven `langchain_azure_ai.agents.hosting.run` entrypoint instead of constructing an `InvocationsHostServer` in application code.
+
+> **No agent code changes are required to migrate an existing LangChain
+> agent to this approach.** Keep the existing graph implementation as-is,
+> point `langgraph.json` at its exported graph, and launch the hosting
+> entrypoint with the desired protocol.
+
+Multi-turn continuity is provided by a LangGraph `MemorySaver`
+checkpointer: the resolved `agent_session_id` is forwarded to the graph
+as `RunnableConfig.configurable.thread_id`, so each session's history is
+preserved in process memory.
+
+## How It Works
+
+### Graph and model integration
+
+The agent uses `langchain_openai.ChatOpenAI` with an Azure bearer token
+provider from `DefaultAzureCredential` and an OpenAI-compatible endpoint
+from `azure.ai.projects.AIProjectClient` (`az login` is enough for local
+development). The graph is the same no-tool `create_agent` graph with
+`MemorySaver` used by the other minimal Invocations samples in this repo.
+
+See [src/langgraph-run-invocations/main.py](src/langgraph-run-invocations/main.py)
+for the graph and [src/langgraph-run-invocations/langgraph.json](src/langgraph-run-invocations/langgraph.json)
+for its entrypoint configuration.
+
+### Migrating an Existing Agent
+
+An existing LangChain agent that exports a compiled graph does not need a
+hosting wrapper, an `InvocationsHostServer` import, or a new `main()`
+function. Add a `langgraph.json` file that references the existing graph
+symbol, then use the command shown below. The agent's model, tools,
+prompts, state, checkpointer, and graph code remain unchanged.
+
+### Agent Hosting
+
+The `langgraph.json` file maps the `agent` name to the compiled graph in
+`main.py`. The hosting entrypoint loads that graph and exposes it through
+the protocol selected by the required `--protocol` argument. This sample
+selects `invocations` in both its local command and Docker image.
+
+## Option 1: Azure Developer CLI (`azd`)
+
+### Prerequisites
+
+1. **Azure Developer CLI (`azd`)** — [Install azd](https://learn.microsoft.com/en-us/azure/developer/azure-developer-cli/install-azd)
+2. Install the Foundry extension:
+
+   ```bash
+   azd ext install microsoft.foundry
+   ```
+
+3. Authenticate:
+
+   ```bash
+   azd auth login
+   ```
+
+### Initialize the agent project
+
+No cloning required. Create a new folder and initialize from the manifest:
+
+```bash
+mkdir hosted-langgraph-agent && cd hosted-langgraph-agent
+azd ai agent init -m https://github.com/microsoft-foundry/foundry-samples/blob/main/samples/python/hosted-agents/langgraph/invocations/03-run/azure.yaml
+```
+
+Follow the prompts to configure your Foundry project and model deployment.
+
+### Provision Azure resources (if needed)
+
+If you don't already have a Foundry project and model deployment:
+
+```bash
+azd provision
+```
+
+### Run the agent locally
+
+```bash
+azd ai agent run --no-client
+```
+
+The agent host will start on `http://localhost:8088`.
+
+### Invoke the local agent
+
+In a separate terminal, invoke the running agent:
+
+```bash
+azd ai agent invoke --local '{"message": "Hello!"}'
+```
+
+Or invoke directly with curl. The returned `x-agent-session-id` header is used for multi-turn conversations:
+
+```bash
+curl -i -X POST http://localhost:8088/invocations \
+    -H "Content-Type: application/json" \
+    -d '{"message": "Hello!"}'
+```
+
+### Deploy to Foundry
+
+Once tested locally, deploy to Microsoft Foundry:
+
+```bash
+azd deploy
+```
+
+## Option 2: VS Code (Foundry Toolkit)
+
+### Prerequisites
+
+1. **VS Code** with the **[Foundry Toolkit](https://marketplace.visualstudio.com/items?itemName=ms-windows-ai-studio.windows-ai-studio)** extension installed.
+2. Install the **[Python](https://marketplace.visualstudio.com/items?itemName=ms-python.python)** extension pack if you want to debug the sample locally.
+
+### Set up the Python virtual environment
+
+- Open the Command Palette (`Ctrl+Shift+P`) and run **Python: Create Environment...** to create a virtual environment in the project.
+- Ensure `pip` is version 26.1 or newer (check with `pip --version`). Older versions fail to resolve this sample's dependencies. Upgrade if needed:
+
+  ```bash
+  python -m pip install --upgrade pip
+  ```
+
+- Install dependencies in the virtual environment:
+
+  ```bash
+  pip install -r src/langgraph-run-invocations/requirements.txt
+  ```
+
+### Run locally
+
+```bash
+cd src/langgraph-run-invocations
+python -m langchain_azure_ai.agents.hosting.run --protocol invocations
+```
+
+Then invoke it:
+
+```bash
+curl -X POST http://127.0.0.1:8088/invocations \
+  -H "Content-Type: application/json" \
+  -d '{"message": "Hello!"}'
+```
+
+## Troubleshooting
+
+### Azure OpenAI Permission Denied (401)
+
+This sample calls the project's model endpoint through LangChain. The hosted agent's managed identity needs the **Foundry User** role on the project. Assign that role at the project scope and retry after role propagation completes.
@@ -0,0 +1,50 @@
+# yaml-language-server: $schema=https://raw.githubusercontent.com/Azure/azure-dev/main/schemas/v1.0/azure.yaml.json
+
+requiredVersions:
+  extensions:
+    azure.ai.agents: '>=1.0.0-beta.4'
+name: langgraph-run-invocations
+services:
+  ai-project:
+    host: azure.ai.project
+    deployments:
+      - name: gpt-5.4-mini
+        model:
+          format: OpenAI
+          name: gpt-5.4-mini
+          version: '2026-03-17'
+        sku:
+          name: GlobalStandard
+          capacity: 10
+  langgraph-run-invocations:
+    host: azure.ai.agent
+    metadata:
+      tags:
+        - AI Agent Hosting
+        - Invocations Protocol
+        - LangGraph
+        - LangChain
+        - Tool Calling
+        - Python
+    project: src/langgraph-run-invocations
+    language: python
+    codeConfiguration:
+      runtime: python_3_13
+      entryPoint: '-m langchain_azure_ai.agents.hosting.run --protocol invocations'
+    uses:
+      - ai-project
+    kind: hosted
+    name: langgraph-run-invocations
+    displayName: LangGraph Run (Invocations)
+    description: 'A minimal configuration-driven LangGraph agent hosted on Foundry over the Invocations protocol using the langchain_azure_ai.agents.hosting.run entrypoint.'
+    protocols:
+      - protocol: invocations
+        version: 2.0.0
+    env:
+      AZURE_AI_MODEL_DEPLOYMENT_NAME: ${AZURE_AI_MODEL_DEPLOYMENT_NAME}
+    container:
+      resources:
+        cpu: '0.5'
+        memory: 1Gi
+infra:
+  provider: microsoft.foundry
@@ -0,0 +1,7 @@
+.venv
+__pycache__
+*.pyc
+*.pyo
+*.pyd
+.Python
+.env
@@ -0,0 +1,6 @@
+# Foundry project endpoint — auto-injected in hosted containers.
+# Only set manually if running without `azd ai agent run`.
+FOUNDRY_PROJECT_ENDPOINT=
+
+# Model deployment name — must match a deployment in your Foundry project.
+AZURE_AI_MODEL_DEPLOYMENT_NAME=
@@ -0,0 +1,17 @@
+FROM python:3.12-slim
+
+WORKDIR /app
+
+COPY . user_agent/
+WORKDIR /app/user_agent
+
+RUN python -m pip install --upgrade pip && \
+    if [ -f requirements.txt ]; then \
+        python -m pip install -r requirements.txt; \
+    else \
+        echo "No requirements.txt found"; \
+    fi
+
+EXPOSE 8088
+
+CMD ["python", "-m", "langchain_azure_ai.agents.hosting.run", "--protocol", "invocations"]
@@ -0,0 +1,8 @@
+{
+  "$schema": "https://langgra.ph/schema.json",
+  "dependencies": ["."],
+  "graphs": {
+    "agent": "./main.py:graph"
+  },
+  "env": ".env"
+}
@@ -0,0 +1,76 @@
+"""Sample 03 - configuration-driven Invocations API host.
+
+Defines the same ``create_agent`` graph with ``MemorySaver`` as sample 01.
+The ``langchain_azure_ai.agents.hosting.run`` module loads the graph from
+``langgraph.json`` and hosts it with the selected protocol.
+
+Required environment variables (set in `.env` or your shell):
+
+    FOUNDRY_PROJECT_ENDPOINT        e.g. https://<acct>.services.ai.azure.com/api/projects/<proj>
+    AZURE_AI_MODEL_DEPLOYMENT_NAME  e.g. gpt-5.4-mini (defaults to "gpt-5.4-mini")
+    PORT                            optional, defaults to 8088
+
+Run::
+
+    az login
+    cp .env.example .env  # then edit the values
+    python -m langchain_azure_ai.agents.hosting.run --protocol invocations
+
+Then in another terminal:
+
+    curl -i -X POST http://127.0.0.1:8088/invocations \
+      -H 'Content-Type: application/json' \
+      -d '{"message":"My name is Alice."}'
+"""
+from __future__ import annotations
+
+import os
+
+from azure.ai.projects import AIProjectClient
+from azure.identity import DefaultAzureCredential, get_bearer_token_provider
+from dotenv import load_dotenv
+from langchain.agents import create_agent
+from langchain_openai import ChatOpenAI
+from langgraph.checkpoint.memory import MemorySaver
+
+from opentelemetry import trace
+from opentelemetry.exporter.otlp.proto.http.trace_exporter import OTLPSpanExporter
+from opentelemetry.sdk.trace import TracerProvider
+from opentelemetry.sdk.trace.export import BatchSpanProcessor
+
+from langchain_azure_ai.callbacks.tracers import enable_auto_tracing
+
+load_dotenv()
+
+_AZURE_AI_SCOPE = "https://ai.azure.com/.default"
+
+
+def _build_chat_model() -> ChatOpenAI:
+    project_endpoint = os.environ["FOUNDRY_PROJECT_ENDPOINT"].rstrip("/")
+    deployment = os.environ.get("AZURE_AI_MODEL_DEPLOYMENT_NAME") or "gpt-5.4-mini"
+    credential = DefaultAzureCredential()
+    project = AIProjectClient(endpoint=project_endpoint, credential=credential)
+    openai_client = project.get_openai_client()
+    token_provider = get_bearer_token_provider(credential, _AZURE_AI_SCOPE)
+
+    return ChatOpenAI(
+        model=deployment,
+        base_url=str(openai_client.base_url),
+        api_key=token_provider,
+        use_responses_api=True,
+        output_version="responses/v1",
+    )
+
+
+def _configure_tracing() -> None:
+    if os.environ.get("OTEL_EXPORTER_OTLP_ENDPOINT"):
+        provider = TracerProvider()
+        provider.add_span_processor(BatchSpanProcessor(OTLPSpanExporter()))
+        trace.set_tracer_provider(provider)
+        enable_auto_tracing()
+    else:
+        enable_auto_tracing(auto_configure_azure_monitor=True)
+
+
+_configure_tracing()
+graph = create_agent(_build_chat_model(), tools=[], checkpointer=MemorySaver())
-Original file line number
+Diff line change
@@ @@ -0,0 +1,7 @@ @@
 +.venv
 +__pycache__
 +*.pyc
 +*.pyo
 +*.pyd
 +.Python
 +.env