Craft Powerful Agents & Workflows in Minutes with fast-agent: The MCP-Native Framework

Tired of wrestling with complex code to build effective agents? fast-agent empowers you to create sophisticated agents and workflows with minimal boilerplate. Leverage its intuitive, declarative syntax and MCP-native support to focus on what matters: crafting killer prompts and orchestrating interactions.

Why Choose fast-agent for Agent Application Development?

• Rapid Development: Build and test agents within minutes using simple, declarative files. No more endless configuration headaches!
• MCP Native: fast-agent is the first framework with complete, end-to-end tested MCP feature support, including Sampling. Unleash the full power of MCP.
• Multi-Modal Mastery: Seamlessly handle images and PDFs with both Anthropic and OpenAI models via prompts, resources, and MCP tool call results. Deliver rich, contextual experiences.

Supported Models:

• Anthropic (Haiku, Sonnet, Opus)
• OpenAI (gpt-4o/gpt-4.1 family, o1/o3 family)

Get Started Now: Your fast-agent Quickstart Guide

Ready to dive in? Follow these simple steps to create your first agent:

1. Install uv: uv pip install fast-agent-mcp
2. Set up example agent and config files: fast-agent setup
3. Run your agent: uv run agent.py
4. Specify a model: uv run agent.py --model=o3-mini.low
5. Explore workflow examples: fast-agent quickstart workflow (for building effective agents examples)

Looking for more complex examples? Explore the included Researcher Agent (with Evaluator-Optimizer workflow) and Data Analysis Agent (similar to the ChatGPT experience), demonstrating MCP Roots support.

Windows Users: Don't forget to check the configuration files for necessary changes related to Filesystem and Docker MCP Servers.

Unleash Agent Potential: Chaining, Parallel Processing & Human Input

fast-agent goes beyond basic agents, offering powerful workflow capabilities. Combine agents, leverage MCP servers, and even request human input for complex tasks.

Workflow Types:

• Chain: Call agents in sequence for multi-step processes.
• Parallel: Send simultaneous messages to multiple agents for diverse perspectives. (fan-out/fan-in)
• Human Input: Request user input during agent execution for enhanced context.
• Evaluator-Optimizer: Refine content iteratively by combining a content generator and quality evaluator agent.
• Router: Direct messages to the most appropriate agent based on content analysis.
• Orchestrator: Create plans to divide complex tasks among available agents.

Building Effective Agents: A Code Example

import asyncio
from mcp_agent.core.fastagent import FastAgent

# Create the application
fast = FastAgent("Agent Example")

@fast.agent(
    instruction="Given an object, respond only with an estimate of its size."
)
async def main():
    async with fast.run() as agent:
        await agent.interactive()

if __name__ == "__main__":
    asyncio.run(main())

Seamless Multimodal Integration

Leverage fast-agent's multimodal capabilities to build agents that can process and understand images and PDFs.

Example: Summarizing a PDF

summary: str = await agent.with_resource(
    "Summarise this PDF please",
    "mcp_server",
    "resource://fast-agent/sample.pdf",
)

Enhance Your Agent Experience: MCP Prompts and Sampling

fine-tune agent behavior by utilizing MCP prompts and sampling techniques.

Advanced Agent Definition Options

Customize your agents with a wealth of options:

• Model selection
• MCP Server integration
• Chat history management
• Human input requests

Contributing to fast-agent

fast-agent is designed to create and interact with sophisticated agents and workflows in minutes. This MCP Native Project invites contributions and PRs to make building effective agents even better.