Supercharge Your Agents: A Deep Dive into Fast-Agent

Tired of wrestling with complex agent frameworks? Fast-agent simplifies creation and interaction with sophisticated Agents and Workflows. Built with simplicity and power in mind, it offers end-to-end tested MCP feature support and seamless integration with leading models like Anthropic and OpenAI. Discover how fast-agent boosts your agent development.

Why Choose Fast-Agent?

• Rapid Development: Declarative syntax lets you focus on crafting prompts and MCP servers.
• Multi-Modal Magic: Supports both Images and PDFs for Anthropic and OpenAI via Prompts, Resources, and MCP Tool Call results.
• Seamless Testing: Passthrough and playback LLMs enable faster Python glue-code development and robust testing.
• MCP Native: Fully integrates with MCP (Microservice Communication Protocol), unlocking advanced agent capabilities.
• Simplified Management: Agent applications are stored in simple files with minimal code, making version control a breeze.

Getting Started with Your First Agent

Jumpstart your journey with these easy steps.

1. Install: uv pip install fast-agent-mcp
2. Setup: fast-agent setup
3. Run: uv run agent.py (or uv run agent.py --model=o3-mini.low to specify a model)
4. Explore: fast-agent quickstart workflow for more examples. These examples include a researcher agent with evaluator-optimizer workflow and a data analysis agent similar to the ChatGPT experience demonstrating MCP roots support. Windows users may need to adjust Filesystem and Docker MCP Servers configurations, which are detailed within the configuration files.

Unleash the Power of Basic Agents

Defining a basic agent is incredibly straightforward:

@ fast. agent (
 instruction = "Given an object, respond only with an estimate of its size."
)

Interact with your agent:

async with fast. run () as agent:
 moon_size = await agent ( "the moon")
 print ( moon_size)

Or start an interactive chat session:

async with fast. run () as agent:
 await agent. interactive ()

Compose Powerful Workflows with Chaining

Fast-agent allows you to chain agents together, creating complex workflows controlled via MCP servers.

@ fast. agent (
 "url_fetcher",
 "Given a URL, provide a complete and comprehensive summary",
 servers = [ "fetch"], # Name of an MCP Server defined in fastagent.config.yaml
)

@ fast. agent (
 "social_media",
 """
 Write a 280 character social media post for any given text.
 Respond only with the post, never use hashtags.
 """,
)

@ fast. chain (
 name = "post_writer",
 sequence = [ "url_fetcher", "social_media"],
)

Advanced Workflow Capabilities

Fast-agent provides a rich set of workflow features for building sophisticated agent applications.

• Human Input: Agents can request human input for context.
• Parallel Processing: Distribute tasks across multiple agents simultaneously for faster results using the parallel workflow.
• Evaluator-Optimizer: Refine content iteratively using generator and evaluator agents, useful for research and content creation tasks. This workflow is extremely valuable for producing high quality automated agents.
• Router: Dynamically route messages to the most appropriate agent based on content.
• Orchestrator: Plan and execute complex tasks by dividing them among available agents.

Deep Dive into Agent Features with Fast-Agent

Fast-agent offers comprehensive customization options for tailoring agent behavior.

• Basic Agent: Configure name, instructions, MCP servers, model, history, params and ability to request human input.
• Chain: Set execution order and message accumulation behavior.
• Parallel: Specify agent fan-out, result aggregation, and request inclusion.
• Evaluator-Optimizer: Define generator, evaluator, quality threshold, and refinement limits.
• Router: Configure agent list, routing model and history.
• Orchestrator: Choose planning approach, agent list, and iterations.

Multimodal Magic: Beyond Text with Fast-Agent

Leverage the power of multimodal support with these key features.

• Resource Integration: Seamlessly incorporate resources like PDFs to drive agent actions.
• MCP Tool Result Conversion: Automatically convert image and embedded resources into user messages to bypass LLM API limitations.
• MCP Prompts: Apply prompts for fine-tuning and adaptive agent responses.

Elevate Agent Performance with Sampling

Optimize your agents further by configuring sampling LLMs per client/server pair within fastagent.config.yaml. Use powerful sampling models for greater result diversity.

Unlock the Potential of Interactive Shell

Test your agents in real-time with the intuitive interactive shell. Use slash commands to further customize the interactive testing.

Ready to supercharge your agent development? With fast-agent, you can build powerful, efficient, and intelligent agents in minutes. Start building effective agents today!