Building an Agentic AI Assistant with LangGraph and MCP

Description and use

An agentic assistant built with LangGraph and MCP orchestrates tool calls, supports multi-turn conversations, and enforces sandboxed file access.

Problem

Create an AI assistant that can perform multi-step tasks by selecting and executing tools, rather than just generating text responses.

Outcome

The assistant can perform multi-step tasks, maintain conversation context, and safely execute file operations within a sandbox, demonstrating the benefits of combining LangGraph, MCP, and tool orchestration.

Workflow

• Interpret user request
• Select appropriate tool via MCP discovery
• Execute tool call (web-search or filesystem)
• Receive and process tool output
• Decide if further actions are needed
• Return final response to user

Evidence

The assistant supports multi-turn conversations and can use a sandboxed filesystem tool and a web-search tool.

Source

The Model Context Protocol (MCP) defines a standard way for applications to communicate with tool-providing servers and discover available capabilities.

Source

My filesystem tool restricts access to a designated sandbox directory and resolves requested paths against that directory.

Source