Conductor AI Agent Examples (.NET)

August 7, 2026 · View on GitHub

176 self-contained example projects for the conductor-ai agent layer. Each directory is its own .csproj with a single Program.cs, and a header comment listing the environment variables it needs.

Running one

export CONDUCTOR_SERVER_URL=http://localhost:8080/api
export CONDUCTOR_AGENT_LLM_MODEL=anthropic/claude-sonnet-4-6

dotnet run --project Conductor.AI.Examples/01_BasicAgent

CONDUCTOR_AGENT_LLM_MODEL is a convention of these examples (see Shared/Settings.cs), not something the SDK itself reads — an Agent takes its model from the Model property.

Most examples need a server with a configured LLM provider; the provider is configured on the server, not in your .NET process. See ../docs/agents/getting-started.md and ../docs/server-setup.md.

The Build agent examples CI job compiles every project in this directory, so they stay buildable. CI does not execute them.

Layout

PrefixCountCovers
01115110The native agent API
Adk*36Google ADK adapter
Sk*20Semantic Kernel adapter
OpenAi*10OpenAI Agents adapter
Shared/Settings.cs, linked into every project

Numbering is roughly thematic and increases in complexity. Suffixed variants (02a, 16b, 63c) are narrower cuts of the base example.

Where to start

GoalExample
Simplest possible agent01_BasicAgent
Tools via [Tool] methods02a_SimpleTools, 02b_MultiStepTools
Typed results03_StructuredOutput
Server-side HTTP / MCP tools04_HttpTools, 04_McpWeather
Multi-agent delegation05_Handoffs, 06_SequentialPipeline, 07_Parallel*
Guardrails21_RegexGuardrails (server regex), 36_SimpleGuardrails (custom worker), 22_LlmGuardrails (server LLM)
Human approval09_HumanInTheLoop, 78_ApprovalWorkflow
Streaming11_Streaming, 76_WaitForMessageStreaming
Credentials16_Credentials and its 16b16h variants
Deploy / serve / run-by-name63b_Serve, 63c_RunByName, 63d_ServeFromAssembly
Stateful agents and message queues51b_StatefulAgentWithWaitForMessage, 75_WaitForMessage, 83_StatefulResume
PLAN_EXECUTE48_Planner, 108_PlanExecuteRefs, 115_PlanExecutePlannerContext
Observability26_OpenTelemetryTracing, 80_LiveDashboard
Skills91_Skills
Schedules92_ScheduledAgent
Scheduler pause/resume verb fallback93_SchedulerVerbFallback

Examples with extra prerequisites

Some need more than a server and a model:

ExampleAlso needs
04_McpWeather, 04_HttpAndMcpToolsAn MCP server (CI uses mcp-testkit on port 3001)
16*_Credentials*A stored secret, and a server that delivers runtimeMetadata
24_CodeExecutionCode execution enabled server-side
77_KafkaConsumerAgentA Kafka broker with the conductor_topic topic
60*/61_GithubCodingAgent*A GitHub token
09*, 32_HumanGuardrail, 78_ApprovalWorkflowInteractive input — they read from stdin

The HITL examples read stdin, so piping works for non-interactive runs:

printf 'y\n' | dotnet run --project Conductor.AI.Examples/09_HumanInTheLoop
  • ../docs/agents/README.md — the agent documentation index
  • ../docs/examples.md — where every example and test suite lives
  • ../Conductor.AI.E2eTests/ — the same features as assertions, run in CI against a live server