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
| Prefix | Count | Covers |
|---|---|---|
01–115 | 110 | The native agent API |
Adk* | 36 | Google ADK adapter |
Sk* | 20 | Semantic Kernel adapter |
OpenAi* | 10 | OpenAI 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
| Goal | Example |
|---|---|
| Simplest possible agent | 01_BasicAgent |
Tools via [Tool] methods | 02a_SimpleTools, 02b_MultiStepTools |
| Typed results | 03_StructuredOutput |
| Server-side HTTP / MCP tools | 04_HttpTools, 04_McpWeather |
| Multi-agent delegation | 05_Handoffs, 06_SequentialPipeline, 07_Parallel* |
| Guardrails | 21_RegexGuardrails (server regex), 36_SimpleGuardrails (custom worker), 22_LlmGuardrails (server LLM) |
| Human approval | 09_HumanInTheLoop, 78_ApprovalWorkflow |
| Streaming | 11_Streaming, 76_WaitForMessageStreaming |
| Credentials | 16_Credentials and its 16b–16h variants |
| Deploy / serve / run-by-name | 63b_Serve, 63c_RunByName, 63d_ServeFromAssembly |
| Stateful agents and message queues | 51b_StatefulAgentWithWaitForMessage, 75_WaitForMessage, 83_StatefulResume |
| PLAN_EXECUTE | 48_Planner, 108_PlanExecuteRefs, 115_PlanExecutePlannerContext |
| Observability | 26_OpenTelemetryTracing, 80_LiveDashboard |
| Skills | 91_Skills |
| Schedules | 92_ScheduledAgent |
| Scheduler pause/resume verb fallback | 93_SchedulerVerbFallback |
Examples with extra prerequisites
Some need more than a server and a model:
| Example | Also needs |
|---|---|
04_McpWeather, 04_HttpAndMcpTools | An MCP server (CI uses mcp-testkit on port 3001) |
16*_Credentials* | A stored secret, and a server that delivers runtimeMetadata |
24_CodeExecution | Code execution enabled server-side |
77_KafkaConsumerAgent | A Kafka broker with the conductor_topic topic |
60*/61_GithubCodingAgent* | A GitHub token |
09*, 32_HumanGuardrail, 78_ApprovalWorkflow | Interactive 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
Related
- ../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