# Agentkit SDK

> A model-driven Python SDK for building AI agents in just a few lines of code.

## Documentation

- [Introduction](/): Overview of Agentkit SDK
- [Quickstart](/quickstart): Build your first agent in 5 minutes
- [Installation](/installation): Install the SDK and configure providers

## Concepts

### Agents
- [Agent Overview](/concepts/agents/overview): The Agent class and its parameters
- [Agent Loop](/concepts/agents/agent-loop): How the reasoning loop works
- [Invocation Limits](/concepts/agents/invocation-limits): Per-call turn and token caps
- [Hooks](/concepts/agents/hooks): Lifecycle event callbacks
- [Structured Output](/concepts/agents/structured-output): Typed Pydantic responses
- [Agent State](/concepts/agents/state): Shared mutable state for tools
- [Streaming](/concepts/agents/streaming): Token streaming and async iteration

### Tools
- [Tools Overview](/concepts/tools/overview): All tool types
- [Function Tools](/concepts/tools/function-tools): Build tools with @tool decorator
- [MCP Tools](/concepts/tools/mcp-tools): Model Context Protocol integration

### Model Providers
- [Models Overview](/concepts/models/overview): Supported providers
- [Amazon Bedrock](/concepts/models/bedrock): Default provider
- [Anthropic](/concepts/models/anthropic): Claude models
- [OpenAI](/concepts/models/openai): GPT and o-series models
- [Google Gemini](/concepts/models/gemini): Gemini models
- [Ollama](/concepts/models/ollama): Local open-source models
- [LiteLLM](/concepts/models/litellm): 100+ providers via LiteLLM

### Multi-Agent
- [Multi-Agent Overview](/concepts/multiagent/overview): Graph and Swarm patterns
- [Graph](/concepts/multiagent/graph): Deterministic agent orchestration
- [Swarm](/concepts/multiagent/swarm): Autonomous collaborative agents
- [Agent as Tool](/concepts/multiagent/agent-as-tool): Sub-agent delegation; `as_tool(mode="reject"|"queue"|"spawn")` for concurrent calls

### Plugins
- [Plugins](/concepts/plugins): Built-in plugins including AgentSkills and AGENTS.md project context
- [Agent API — AgentConfig](/api-reference/agent#agentconfig): Agent metadata and execution options

### Other
- [Conversation Management](/concepts/conversation-management): Context window management
- [Sessions](/concepts/sessions): Persist conversations across runs

## Guides
- [Building Tools](/guides/building-tools): Best practices for writing tools
- [MCP Integration](/guides/mcp-integration): Connecting MCP servers
- [Deploying](/guides/deploying): FastAPI, Lambda, Docker deployment

## API Reference
- [Agent API](/api-reference/agent): Full Agent class reference
- [Tools API](/api-reference/tools): Tool decorator and MCPClient
- [Models API](/api-reference/models): All model provider classes
- [Multi-Agent API](/api-reference/multiagent): Graph and Swarm classes
- [Hooks API](/api-reference/hooks): All hook events
- [Sessions API](/api-reference/sessions): Session managers and snapshots

## Quick Example

```python
from elsai import Agent, tool

@tool
def get_weather(city: str) -> str:
    """Get current weather for a city.

    Args:
        city: The city name.
    """
    return f"Sunny, 22°C in {city}"

agent = Agent(tools=[get_weather])
result = agent("What's the weather in Tokyo?")
print(result)
```

## Installation

```bash
pip install --extra-index-url https://elsai-agents.elsai.ai/root/ elsai-agents==0.3.0
```

## Links

- GitHub: https://github.com/elsai-agents/sdk-python
- PyPI: https://pypi.org/project/elsai-agents/
- Discord: https://discord.gg/elsai