Fast Agent + MCP — Build Real LLM Agents in Minutes

Expand your AI Engineering toolkit with Fast Agent.

This fun little tool is a CLI and Python library.

I’ll show you how to use it for quickly prototyping LLM applications with MCP across multiple providers.

Check out the video on YouTube

Timestamps:
1:09 - CLI Demo
6:27 - Installation
8:35 - Basic Agent (Demo #1)
13:53 - Parallel LLMs (Demo #2)
19:47 - Workflows (Demo #3)

Massive Course Update 🚀

I’ve added 32 videos to the AI Engineer Roadmap, over 10 hours of content, making this a proper 100% complete video course on the foundational skills needed for AI Engineering.

With this much anticipated (by me, at least—LOL) update, I’ve shipped a complete set of videos lessons for every topic group in the course. I’ve taken a full month off YouTube (5 weeks) in order to ship this for you guys.

Have a look at the first lesson video for free and consider buying the full course if you like it.

I’ve also increased the price of the course to reflect the new value added. I think it’s still an absolute steal at $79 USD.

Topics from this week’s video

• Introduction to Fast Agent

  • Python library and CLI for building and interacting with LLM agents

  • Supports multiple providers and much of the MCP spec

  • Useful for testing MCP tools and model parallelization

• CLI Demonstration

  • Running commands like go with different models (GPT-5 nano, Haiku, etc.)

  • MCP Server Integration

    • Using remote MCP servers (sequential thinking example)

    • Using local MCP servers with uvx

• Installation and Setup

  • Installing Fast Agent via pipx install fast-agent-mcp

  • Recommendation to install inside project environments with uv

  • Setting up Python 3.13 for GPT-5 compatibility

  • Managing virtual environments and ensuring correct Python/Fast Agent version

• Python Library Usage

  • Writing a basic agent with decorators

  • Running agents with uv run vs. explicit Python virtual environments

• Parallelization Demonstration

  • Comparing outputs across multiple models simultaneously

  • Example: prompt sent to GPT-5, Sonnet, and Gemma 3

  • Using results programmatically for follow-up questions and structured outputs

• Workflows Demonstration

  • Generating workflow examples with fast agent quickstart

  • Example: chaining agents (URL fetcher → social media summarizer)

  • Human input integration and tool-calling demonstration

• Advanced Example: D&D Encounter Simulation

  • Creating an agent with access to random number generator MCP server

  • Model rolls dice to simulate actions in a role-playing battle

  • Example interaction: Warlock vs. Vampire encounter

• Workflow Patterns

  • Router agent: chooses which sub-agent to use dynamically

  • Orchestrator agent: more complex workflow coordination

---

If you believe your product or service can fulfill a true need, it is your moral obligation to sell it.

Zig Ziglar