fast-agent-mcp

What it does

fast-agent-mcp is a CLI framework for building, running, and evaluating LLM agents with integrated MCP server support. It provides a unified interface for interacting with multiple LLM providers—Anthropic, OpenAI, Google, Azure, Ollama, Deepseek—alongside MCP servers for tool composition. The framework enables interactive agent sessions with shell access, file system operations, and network calls, along with structured outputs, vision, and PDF support.

Who it's for

Developers building LLM-powered agents or workflows; teams setting up agent evaluation platforms; engineers composing MCP servers for agentic tasks; researchers prototyping multi-model agent configurations.

Common use cases

• Build a coding agent with shell execution and file I/O for automated development tasks
• Run an interactive agent session with MCP server connections to external tools or APIs
• Compose multiple MCP servers into a single agent workflow for complex task execution
• Test agent prompts and MCP interactions across different LLM models to optimize performance
• Set up evaluation frameworks for comparing agent outputs on structured tasks

Setup pitfalls

• 8 secrets detected in source: Review the codebase before production use. Ensure credentials are loaded from environment variables or secure vaults, not embedded.
• High-risk classification: Agents have filesystem write and shell execution capabilities—sandbox execution and validate MCP server endpoints before connecting to untrusted sources.
• Requires uv package manager: Install via astral.sh/uv; standard pip alone is insufficient for some installation paths.
• Model credentials required: Populate API keys for your chosen provider (e.g., ANTHROPIC_API_KEY, OPENAI_API_KEY) before running agents.
• MCP endpoint configuration: Distinguish between stdio (local, spawned by the agent) and HTTP (remote, requires credential storage)—mismatched transport can cause tool failures.