LangGraph Browser Use

Create AI agents that can use web browsers with LangGraph and MarinaBox

LangGraph Integration

MarinaBox can be integrated with LangGraph to create AI agents that can interact with isolated web browsers. This guide shows you how to set up MarinaBox as a Browser Tool for LangGraph agents.

Prerequisites

• MarinaBox installed and configured
• Python 3.8+
• Anthropic API key for Claude

Basic Setup

First, install the required dependencies:

pip install marinabox langgraph langchain-anthropic

Set up your Anthropic API key:

import os
os.environ['ANTHROPIC_API_KEY'] = "your-anthropic-api-key"

Creating the Workflow

Here’s a complete example showing how to create a LangGraph workflow with MarinaBox:

from marinabox import mb_start_browser, mb_stop_browser, mb_use_browser_tool
from langchain_core.messages import AIMessage, HumanMessage
from langchain_core.tools import tool
from langgraph.prebuilt import ToolNode
from langgraph.graph import StateGraph, START, END
from langchain_anthropic import ChatAnthropic
from typing import Annotated, Literal
from langgraph.types import Command

# Set up tools and model
tools = [mb_use_browser_tool]
tool_node = ToolNode(tools=tools)
model_with_tools = ChatAnthropic(
    model="claude-3-5-sonnet-20241022", 
    temperature=0
).bind_tools(tools)

# Define workflow logic
def should_continue(state: Annotated[dict, InjectedState()]):
    messages = state["messages"]
    if len(messages) > 0:
        last_message = messages[-1]
        if last_message.tool_calls:
            return Command(goto="tool_node")
    return Command(goto="stop_browser")

def call_model(state: Annotated[dict, InjectedState()]):
    input_message = input("Enter your message: ")
    if input_message != "stop_browser":
        messages = [HumanMessage(content=input_message)]
        response = model_with_tools.invoke(messages)
        return {
            "messages": [response], 
            "session_id": state.get("session_id")
        }
    return {
        "messages": [], 
        "session_id": state.get("session_id")
    }

# Create workflow
workflow = StateGraph(dict)

# Add nodes
workflow.add_node("start_browser", mb_start_browser)
workflow.add_node("agent", call_model)
workflow.add_node("tool_node", tool_node)
workflow.add_node("stop_browser", mb_stop_browser)
workflow.add_node("should_continue", should_continue)

# Add edges
workflow.add_edge(START, "start_browser")
workflow.add_edge("start_browser", "agent")
workflow.add_edge("tool_node", "agent")
workflow.add_edge("agent", "should_continue")
workflow.add_edge("stop_browser", END)

# Compile and run
app = workflow.compile()
app.invoke({"messages": ""})

MarinaBox provides three key LangGraph components:

• mb_start_browser: A LangGraph node that initializes and starts up an isolated browser session
• mb_stop_browser: A LangGraph node that safely closes the browser session
• mb_use_browser_tool: A LangGraph tool that enables agents to interact with the browser using natural language commands

How It Works

1. Start Browser Node: The workflow begins with the mb_start_browser node, which creates and initializes an isolated browser session
2. Browser Control Tool: The agent uses the mb_use_browser_tool to send natural language commands to control the browser
3. Stop Browser Node: When finished, the mb_stop_browser node ensures proper cleanup and closure of the session

Monitoring Sessions

You can monitor the browser session in real-time using the VNC port provided by MarinaBox. For embedding the live view in your application, see our Embedding Guide.

Example Usage

Here’s a sample interaction with the agent:

Enter your message: Please search for "LangGraph documentation" on Google
# Agent will:
# 1. Navigate to Google
# 2. Search for LangGraph documentation
# 3. Display the results

Enter your message: stop_browser
# This will cleanly close the browser session

---

Get Support & Join Our Community

• Discord: https://discord.gg/nAyFBSSU87
• Email: [e-mail]