microsoft--agent-framework/dotnet/samples/04-hosting/FoundryHostedAgents/responses/Hosted-McpTools

Hosted-McpTools

A hosted agent demonstrating two layers of MCP (Model Context Protocol) tool integration:

1. Client-side MCP (Microsoft Learn) — The agent connects directly to the Microsoft Learn MCP server via McpClient, discovers tools, and handles tool invocations locally within the agent process.

2. Server-side MCP (Microsoft Learn) — The agent declares a HostedMcpServerTool which delegates tool discovery and invocation to the LLM provider (Azure OpenAI Responses API). The provider calls the MCP server on behalf of the agent with no local connection needed.

How the two MCP patterns differ

	Client-side MCP	Server-side MCP
Connection	Agent connects to MCP server directly	LLM provider connects to MCP server
Tool invocation	Handled by the agent process	Handled by the Responses API
Auth	Agent manages credentials	Provider manages credentials
Use case	Custom/private MCP servers, fine-grained control	Public MCP servers, simpler setup
Example	Microsoft Learn (McpClient + HttpClientTransport)	Microsoft Learn (HostedMcpServerTool)

Prerequisites

• .NET 10 SDK
• A Foundry project with a deployed model (e.g., gpt-4o)
• Azure CLI logged in (az login)

Configuration

Copy the template and fill in your values:

cp .env.example .env

Edit .env:

FOUNDRY_PROJECT_ENDPOINT=https://<your-account>.services.ai.azure.com/api/projects/<your-project>
FOUNDRY_MODEL=gpt-4o

Running directly (contributors)

cd dotnet/samples/04-hosting/FoundryHostedAgents/responses/Hosted-McpTools
dotnet run

Test it

Using the Azure Developer CLI:

# Uses GitHub MCP (client-side)
azd ai agent invoke --local "Search for the agent-framework repository on GitHub"

# Uses Microsoft Learn MCP (server-side)
azd ai agent invoke --local "How do I create an Azure storage account using az cli?"

Running with Docker

1. Publish for the container runtime

dotnet publish -c Debug -f net10.0 -r linux-musl-x64 --self-contained false -o out

2. Build and run

docker build -f Dockerfile.contributor -t hosted-mcp-tools .

export AZURE_BEARER_TOKEN=$(az account get-access-token --resource https://ai.azure.com --query accessToken -o tsv)

docker run --rm -p 8088:8088 \
  -e AGENT_NAME=mcp-tools \
  -e AZURE_BEARER_TOKEN=$AZURE_BEARER_TOKEN \
  --env-file .env \
  hosted-mcp-tools

Deploying to Foundry (azd spec)

This sample includes an azd manifest (agent.manifest.yaml) and hosted agent spec (agent.yaml) for deployment to Foundry.

Initialize an azd project from this sample's manifest:

mkdir mcp-tools && cd mcp-tools
azd ai agent init -m https://github.com/microsoft/agent-framework/blob/main/dotnet/samples/04-hosting/FoundryHostedAgents/responses/Hosted-McpTools/agent.manifest.yaml

Then deploy:

azd deploy

If you need to override defaults, set deployment-time environment variables in the azd environment before deploying:

azd env set AGENT_NAME mcp-tools
azd env set AZURE_AI_MODEL_DEPLOYMENT_NAME gpt-4o

For end-to-end hosted agent deployment guidance, see the official deployment guide.

---

NuGet package users

Use the standard Dockerfile instead of Dockerfile.contributor. See the commented section in HostedMcpTools.csproj for the PackageReference alternative.

Related samples

• Hosted-Toolbox/ — connects to a single Foundry Toolbox via the AF Foundry hosting bridge (AddFoundryToolboxes + FoundryAITool.CreateHostedMcpToolbox).
• Hosted-Toolbox-AuthPaths/ — same hosting bones as Hosted-Toolbox/, but the toolbox bundles three MCP tools each authenticated differently (key, Entra agent identity, inline Authorization), driven by the shared Using-Samples/SimpleAgent/ REPL.