> ## Documentation Index
> Fetch the complete documentation index at: https://docs.orxhestra.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Examples

> Runnable Python demos for every feature: basic agents, streaming, multi-agent pipelines, planners, structured output, A2A servers, and YAML composer.

The `examples/` directory contains runnable demos for every major feature. Each example includes a `NotImplementedError` placeholder - replace it with your LLM of choice.

## Core Examples

| Example                 | What it demonstrates                                           |
| ----------------------- | -------------------------------------------------------------- |
| `basic_agent.py`        | Minimal agent with a single tool                               |
| `streaming_agent.py`    | AgentTool sub-agent streaming with real-time event bubbling    |
| `multi_agent.py`        | SequentialAgent pipeline (researcher -> writer)                |
| `parallel_agent.py`     | ParallelAgent running 3 research agents concurrently           |
| `loop_agent.py`         | LoopAgent with writer/reviewer iterative refinement            |
| `transfer_agent.py`     | Agent handoff via `make_transfer_tool` (triage -> specialists) |
| `react_agent.py`        | ReActAgent with structured thought-action-observation loop     |
| `planner_agent.py`      | PlanReActPlanner with structured chain-of-thought              |
| `memory_agent.py`       | Persistent memory with `InMemoryMemoryService`                 |
| `memory_system.py`      | File-based persistent memory with YAML frontmatter             |
| `structured_output.py`  | Typed Pydantic output via `output_schema`                      |
| `human_in_the_loop.py`  | Interactive `ask_human` tool for user input                    |
| `task_orchestration.py` | TaskPlanner with managed task board                            |
| `multi_provider.py`     | Same agent across OpenAI, Anthropic, and Google providers      |
| `database_sessions.py`  | Persistent sessions with SQLite via `DatabaseSessionService`   |
| `background_tasks.py`   | Background task lifecycle with `TaskStore`                     |
| `mcp_agent.py`          | MCP tool server integration                                    |
| `langfuse_tracing.py`   | Langfuse tracing integration                                   |
| `a2a_server.py`         | A2A protocol server endpoint                                   |
| `thinking_parts.py`     | Processing extended thinking/reasoning tokens                  |
| `signed_events.py`      | Cryptographic event signing and verification                   |
| `artifact_tools.py`     | Saving and loading artifacts during agent execution            |
| `tool_confirmation.py`  | Gating dangerous tools with confirmation callbacks             |

```bash theme={null}
# Run any example (after replacing the NotImplementedError with your LLM)
uv run python examples/basic_agent.py
```

## Composer Examples

YAML-based agent composition examples in `examples/composer/`:

| Example                | What it demonstrates                                  |
| ---------------------- | ----------------------------------------------------- |
| `composer/simple/`     | Single agent, minimal YAML config                     |
| `composer/pipeline/`   | Sequential researcher -> writer pipeline with tools   |
| `composer/loop/`       | Draft/review loop with `exit_loop` termination        |
| `composer/transfer/`   | Triage agent with transfer routing to specialists     |
| `composer/coding/`     | Plan-code-review loop with filesystem and shell tools |
| `composer/a2a-server/` | A2A endpoint with Dockerfile for deployment           |

```bash theme={null}
# Run a composer example
uv run python examples/composer/simple/run.py

# Run the A2A server example
uv run python examples/composer/a2a-server/run.py
```