MCP AI Agent

What is MCP AI Agent?

MCP AI Agent is a TypeScript library that enables AI agents to leverage MCP (Model Context Protocol) servers for enhanced capabilities. It integrates with the AI SDK to provide seamless connection to MCP servers, automatically discovering and using their tools in AI-powered applications. Designed for developers, it runs in Node.js/TypeScript environments.

How to use MCP AI Agent?

Install via npm install mcp-ai-agent, then import AIAgent and Servers. Create an agent by passing a configuration object with a name, description, model (e.g., from @ai-sdk/openai), and toolsConfigs array referencing preconfigured servers or custom tool definitions. Call agent.generateResponse({ prompt }) to execute tasks. For multi-agent systems, compose specialized agents into a master agent using type: "agent" entries.

Key features of MCP AI Agent

• Connects to multiple MCP servers via STDIO or SSE transports
• Automatically discovers and exposes tools from MCP servers
• Supports agent composition: specialized agents delegate tasks
• Includes preconfigured servers for quick setup
• Allows custom tool definitions directly in agent configuration
• Provides verbose mode for debugging

Use cases of MCP AI Agent

• Break down complex problems with sequential thinking
• Search the web and store information in persistent memory
• Build a crew of specialized agents for project management workflows
• Create calculators or domain-specific tools without external servers
• Combine multiple MCP servers into a unified AI agent

FAQ from MCP AI Agent

What is MCP AI Agent?

It is a TypeScript library that lets AI agents use MCP servers to gain additional tools and capabilities, integrating with the AI SDK for text generation.

What MCP servers are preconfigured?

The library ships with support for Sequential Thinking, Memory, AWS KB Retrieval, Brave Search, Everart, Fetch, Firecrawl MCP, and SQLite servers.

Can I add custom tools to an agent?

Yes, you can define custom tools using Zod schemas and execute functions within the agent’s toolsConfigs array, as shown in the Custom Tools example.

Do agents support calling other agents?

Yes, multi-agent workflows are supported. You can compose specialized agents into a master agent that delegates tasks to them via type: "agent" entries.

Is automatic Swagger/OpenAPI conversion supported?

Not yet; it is on the roadmap as “Automatic Swagger/OpenAPI to tools conversion (stateless servers simple integration)” but not currently implemented.