FastAgentic

Build agents with anything. Ship them with FastAgentic.

Documentation • Quickstart • Examples

---

The Deployment Layer for AI Agents

Your agent works locally. Now make it production-ready in minutes, not weeks.

FastAgentic wraps any agent framework—PydanticAI, LangGraph, CrewAI, LangChain—and adds everything you need for production:

You Get	Without Rewriting Your Agent
Checkpointing	Resume 90-minute research agents after crashes
Observability	See every step, token, and decision
Cost Control	Token budgets, rate limits, circuit breakers
Security	OAuth, RBAC, PII detection out of the box
Protocols	MCP + A2A support for tool sharing and collaboration

Your Agent (any framework)  →  FastAgentic  →  Production-Ready API

Install

pip install fastagentic

30-Second Example

Wrap any existing agent with zero code changes:

from your_agent import research_agent  # Any agent, any framework
from fastagentic import run_opaque

# One line. That's it.
result = await run_opaque(
    research_agent,
    query="AI trends 2025",
    run_id="research-001",  # Enables resume on crash
)

Re-run with the same run_id? Instant cached result. No re-execution.

Full Application

from fastagentic import App, tool, agent_endpoint
from fastagentic.adapters import LangGraphAdapter

app = App(title="My Agent API")

@tool
async def search(query: str) -> list[str]:
    """Search the knowledge base."""
    return await kb.search(query)

@agent_endpoint("/chat", runnable=LangGraphAdapter(my_graph))
async def chat(message: str) -> str:
    ...

fastagentic run              # HTTP API server
fastagentic mcp serve        # MCP stdio server
fastagentic agent chat       # Interactive testing

Framework Adapters

Framework	Adapter	Install
PydanticAI	PydanticAIAdapter	pip install fastagentic[pydanticai]
LangGraph	LangGraphAdapter	pip install fastagentic[langgraph]
CrewAI	CrewAIAdapter	pip install fastagentic[crewai]
LangChain	LangChainAdapter	pip install fastagentic[langchain]

Why FastAgentic?

The problem: You've built an agent. It works. But shipping it means adding:

• Crash recovery for long-running tasks
• Monitoring and cost tracking
• Authentication and authorization
• Rate limiting and retries
• API endpoints and protocols

The solution: FastAgentic handles all of this. You keep your agent code unchanged.

Documentation

docs.neullabs.com/fastagentic

• Getting Started — Up and running in 5 minutes
• Adapters — Connect any framework
• Checkpointing — Resume long-running agents
• MCP Protocol — Tool sharing standard
• Deployment — Docker, Kubernetes, cloud

Contributing

git clone https://github.com/neullabs/fastagentic
cd fastagentic
uv sync --extra dev
uv run pytest tests/ -v

License

MIT