Browser History

You can run a local MCP server that gives you access to your browser history data for in-depth analysis and insights. It supports multiple browsers, groups sessions by time, categorizes sites, and lets you explore domain trends—all while keeping data processing local to your machine.

How to use

You connect to the Browser History MCP Server using your MCP client. Start the server locally, then use the client tools to health-check the service, verify which browsers are available, retrieve raw history data, and run analyses. Use the core tools to get quick history access, perform productivity analyses, and generate insights reports. You can search history for specific queries, assign uncategorized URLs to categories, and run Safari support diagnostics if needed.

How to install

Prerequisites include Python 3.12 or higher and one of Firefox, Chrome, or Safari. You also need a way to run MCP clients (uv is recommended). Follow these steps to install and start the Browser History MCP Server locally.

Install uv for dependency management if you haven’t already.

curl -LsSf https://astral.sh/uv/install.sh | sh
uv sync

Test locally to ensure everything is wired up.

uv run mcp dev server/main.py

Install the MCP server for Claude Desktop (restart Claude Desktop afterward). Use the following command.

uv run mcp install server/main.py --name "Browser History MCP"

Additional configuration and essentials

Automatic setup detects your browser profiles to start processing without manual path configuration. If automatic detection fails, you can configure paths manually in the server script.

Manual configuration example (edit the server file to set paths):

FIREFOX_PROFILE_DIR = "/path/to/your/firefox/profile"
CHROME_PROFILE_DIR = "/path/to/your/chrome/profile"

Troubleshooting

If you need to adjust the MCP configuration, you can provide an explicit MCP config snippet that the client reads to start the server. The example below shows how to configure the local MCP runner to start the Browser History MCP Server.

{
  "mcpServers": {
    "Browser History MCP": {
      "command": "/usr/local/bin/uv",
      "args": [
        "run",
        "--with",
        "mcp[cli]",
        "mcp",
        "run",
        "[wherever-you-saved-the-repo]/browser-mcp-server/server/main.py"
      ]
    }
  }
}

Privacy & Security

All data processing happens locally on your machine. No browser history data is transmitted to external servers, and the MCP server reads directly from browser SQLite databases. Optional local caching can be enabled for performance.

License

This project is licensed under the MIT License. You are free to use, modify, and distribute it in accordance with the license terms.

Available tools

health_check

Simple health check to test if the MCP server is working

check_browser_status

Check which browsers are available and which are locked

get_browser_history

Get raw browser history data without analysis (fastest)

analyze_browser_history

Main analysis tool with options for quick_summary, basic, or comprehensive analysis

search_browser_history

Search browser history for specific queries

suggest_categories

Get uncategorized URLs for custom categorization

diagnose_safari_support

Safari support and accessibility diagnostics

productivity_analysis

Comprehensive productivity assessment with metrics and recommendations

learning_analysis

Deep learning pattern analysis for insights into educational content consumption

research_topic_extraction

Extract and summarize research topics from browsing activity

generate_insights_report

Create personalized browsing insights report with highlights and trends

compare_time_periods

Compare browsing habits across time for trend analysis