Skip to main content
MCPAgent runs through LangChain chat models. You can either pass a "provider/model" string and let the agent create the model, or pass an existing LangChain chat model instance. The model must support tool calling. Structured output and streaming are optional, but they are required for the corresponding MCPAgent features to work well.

​
Use simplified mode

Simplified mode supports openai, anthropic, google, and groq provider strings. Install the matching LangChain package and set the provider API key in the environment.
Provider stringPackageAPI key
openai/model@langchain/openaiOPENAI_API_KEY
anthropic/model@langchain/anthropicANTHROPIC_API_KEY
google/model@langchain/google-genaiGOOGLE_API_KEY or GOOGLE_GENERATIVE_AI_API_KEY
groq/model@langchain/groqGROQ_API_KEY
import { MCPAgent } from "mcp-use";

const agent = new MCPAgent({
  llm: "openai/gpt-4o",
  llmConfig: {
    temperature: 0.2,
    maxTokens: 1000,
  },
  mcpServers: {
    filesystem: {
      command: "npx",
      args: ["-y", "@modelcontextprotocol/server-filesystem", "./"],
    },
  },
});

const result = await agent.run({
  prompt: "Summarize the files in this directory.",
});

console.log(result);
await agent.close();
Pass llmConfig.apiKey when you cannot use environment variables. 
const agent = new MCPAgent({
  llm: "anthropic/claude-sonnet-4-6",
  llmConfig: {
    apiKey: process.env.ANTHROPIC_API_KEY,
    temperature: 0,
  },
  mcpServers,
});

​
Use explicit mode

Explicit mode accepts any LangChain-compatible chat model. Use it for custom model setup, provider-specific options, callbacks, or models not supported by simplified mode.

​
OpenAI

import { ChatOpenAI } from "@langchain/openai";
import { MCPAgent, MCPClient } from "mcp-use";

const client = new MCPClient({ mcpServers });
const llm = new ChatOpenAI({
  model: "gpt-4o",
  temperature: 0.7,
  apiKey: process.env.OPENAI_API_KEY,
});

const agent = new MCPAgent({ llm, client });

​
Anthropic

import { ChatAnthropic } from "@langchain/anthropic";
import { MCPAgent, MCPClient } from "mcp-use";

const client = new MCPClient({ mcpServers });
const llm = new ChatAnthropic({
  model: "claude-sonnet-4-6",
  temperature: 0.7,
  apiKey: process.env.ANTHROPIC_API_KEY,
});

const agent = new MCPAgent({ llm, client });

​
Google Gemini

import { ChatGoogleGenerativeAI } from "@langchain/google-genai";
import { MCPAgent, MCPClient } from "mcp-use";

const client = new MCPClient({ mcpServers });
const llm = new ChatGoogleGenerativeAI({
  model: "gemini-pro",
  temperature: 0.7,
  apiKey: process.env.GOOGLE_API_KEY,
});

const agent = new MCPAgent({ llm, client });

​
Groq

import { ChatGroq } from "@langchain/groq";
import { MCPAgent, MCPClient } from "mcp-use";

const client = new MCPClient({ mcpServers });
const llm = new ChatGroq({
  model: "llama-3.1-70b-versatile",
  temperature: 0.7,
  apiKey: process.env.GROQ_API_KEY,
});

const agent = new MCPAgent({ llm, client });

​
Verify provider support

Before wiring a model into a production workflow, run one prompt that requires a tool call. If the model cannot call tools, the agent will not be able to use MCP tools even if the client connection succeeds. For more providers, use a LangChain chat model that supports tool calling and pass it through explicit mode. See the LangChain JavaScript chat model integrations.

​
Next steps