Skip to content

AgentiCraft Documentation

Welcome to the AgentiCraft documentation! AgentiCraft is a production-ready framework for building AI agents with transparent reasoning, streaming capabilities, and comprehensive observability.

🚀 What's New in v0.2.0-alpha

🌊 Streaming Support

Real-time token-by-token responses for all providers: 

async for chunk in agent.stream("Tell me a story"):
    print(chunk.content, end="", flush=True)

🧠 Advanced Reasoning Patterns

Three sophisticated reasoning patterns that make agent thinking transparent: - Chain of Thought: Step-by-step reasoning with confidence tracking - Tree of Thoughts: Multi-path exploration for creative solutions - ReAct: Combines reasoning with tool actions

from agenticraft.agents.reasoning import ReasoningAgent

agent = ReasoningAgent(reasoning_pattern="chain_of_thought")
response = await agent.think_and_act("Solve this complex problem")

# See the reasoning process
for step in response.reasoning_steps:
    print(f"{step.number}. {step.description} (confidence: {step.confidence:.0%})")

🔌 Model Context Protocol (MCP)

Seamless integration with Anthropic's MCP ecosystem: 

from agenticraft.protocols.mcp import MCPServer, MCPClient

# Use MCP tools in your agents
client = MCPClient("ws://localhost:8765")
agent = Agent(tools=[client.get_tool("calculator")])

📊 Production Telemetry

Built-in OpenTelemetry support with <1% overhead: 

from agenticraft.telemetry import setup_telemetry

setup_telemetry(
    service_name="my-agent-service",
    otlp_endpoint="http://localhost:4318",
    enable_metrics=True,
    enable_tracing=True
)

💾 Advanced Memory Systems

Vector and knowledge graph memory for intelligent context: 

from agenticraft.memory import VectorMemory, KnowledgeGraphMemory

# Semantic search across conversations
memory = VectorMemory()
relevant_context = await memory.search("previous discussions about AI")

See all v0.2.0-alpha features →

📚 Documentation Structure

Getting Started

Features

API Reference

Migration Guides

Quick Reference

Examples

Guides

🚀 Key Features

Dynamic Provider Switching

Switch between OpenAI, Anthropic, and Ollama at runtime:

agent.set_provider("anthropic", model="claude-3-opus-20240229")
response = await agent.run("Complex task requiring powerful model")

agent.set_provider("ollama", model="llama2")
response = await agent.run("Simple task that can use local model")

Learn more →

Streaming Responses

Real-time, token-by-token output with visual progress:

# With progress bar
async for chunk in agent.stream_with_progress("Generate a report"):
    # Automatic progress visualization
    pass

Learn more →

Advanced Reasoning

Make agent thinking transparent with structured reasoning patterns:

# Automatic pattern selection
agent = ReasoningAgent(reasoning_pattern="auto")
response = await agent.think_and_act(query)

Learn more →

Production Observability

Built-in telemetry for monitoring, debugging, and optimization:

# Automatic tracing of all operations
with tracer.start_as_current_span("complex_workflow"):
    response = await agent.run("Process customer request")

Learn more →

📖 Start Here

New to AgentiCraft? Start with these resources:

  1. Quick Start Guide - Get up and running in 5 minutes
  2. Reasoning Patterns Guide - Learn about transparent reasoning
  3. Streaming Guide - Real-time responses
  4. Examples - 50+ working examples

By Use Case

Building a chatbot? - Start with Streaming Responses - Add Memory Systems - Deploy with Telemetry

Creating an autonomous agent? - Use Advanced Reasoning - Design with Enhanced Workflows - Monitor with Observability

Building tool integrations? - Explore MCP Protocol - Create Custom Tools - Share via Plugin Marketplace

🔍 How to Use This Documentation

  • Feature Guides: In-depth explanations of each feature with examples
  • API Reference: Detailed technical documentation of all classes and methods
  • Migration Guides: Step-by-step instructions for upgrading
  • Quick Reference: Concise syntax and common patterns
  • Examples: Working code you can run and modify

💡 Getting Help

  • Discord: Join our community Discord
  • GitHub Issues: Report bugs or request features
  • Stack Overflow: Tag questions with agenticraft

🤝 Contributing

We welcome contributions! See our Contributing Guide to get started.

AgentiCraft - Dead simple AI agents with reasoning traces