BasicAgent

Basic Agent: Single-Process Durable Conversation

Overview

The simplest way to get started with Temporalio.Extensions.Agents. A single process registers a Temporal worker and an AI agent, then runs a multi-turn conversation where every exchange is durably recorded in workflow state.

This sample demonstrates:

• Registering an agent with AddTemporalAgents() + opts.AddDurableAgent("name", agent => { ... })
• Resolving the agent's IChatClient via agent.ChatClient = sp => ...
• Adding a tool with agent.AddTool(weatherTool)
• Opening a session with CreateSessionAsync() and sending turns with RunAsync()
• Multi-turn context: the caller sends only the new message — the library retrieves full conversation history from workflow state and includes it in each LLM call automatically

Highlights

• WorkflowUpdate as request/response. Each RunAsync() call is a Temporal [WorkflowUpdate] — the caller blocks until the agent responds, with no polling loop required.
• History lives in the workflow. Conversation context is stored in AgentWorkflow state, not in the client. Any process with the session ID can send a follow-up.
• Sessions are lazy. CreateSessionAsync() creates a local session token — no Temporal workflow exists yet. The AgentWorkflow starts on the first RunAsync() call. Querying the workflow handle before sending any message will return "workflow not found".
• Tool calls are durable per-tool activities. get_weather executes inside its own InvokeAgentTool activity. If the worker restarts between the tool call and the model response, the tool result is replayed from history.
• Single-process simplicity. Worker, agent, and caller all live in the same IHost — the minimum viable setup before splitting into separate processes.

Getting Started

Prerequisites

• .NET 10 SDK or later
• A local Temporal server: temporal server start-dev
• An OpenAI-compatible API key

Configure API credentials

dotnet user-secrets set "OPENAI_API_KEY" "sk-..." --project samples/MAF/BasicAgent
dotnet user-secrets set "OPENAI_API_BASE_URL" "https://api.openai.com/v1" --project samples/MAF/BasicAgent

Run

dotnet run --project samples/MAF/BasicAgent/BasicAgent.csproj

Expected Output

Worker started. Sending messages...

Session workflow ID: ta-assistant-<guid>

User : What is the capital of France?
Agent: The capital of France is Paris.

User : What is its population?
Agent: Paris has a population of approximately 2.1 million in the city proper...

User : What is the weather like there right now?
Agent: The current weather in Paris is sunny. (result is random — get_weather returns "It's sunny" or "It's raining")

Done.