search-history
MCP sever to give AI access to your browser history
claude mcp add --transport stdio ronantakizawa-search-history-mcp uvx claude-desktop
How to use
This MCP server provides tools to access and search your local browser history through an LLM chat interface. It includes three tools: search_history, which lets you search for terms within URLs and page titles; get_recent_history, which retrieves the most recently visited sites; and get_most_visited, which returns your most frequently visited domains. Use these tools to feed contextual history to your chat prompts, personalize LLM responses, or generate a profile based on your browsing activity. The server runs locally and reads from your browser history databases in read-only mode, creating temporary copies to avoid locking issues. No data leaves your machine; results are returned to your local LLM chat app through the MCP interface.
To interact, invoke the available tools via the MCP interface in your chat application or development environment. For example, you can query search_history with a term like “research tools” to surface relevant pages from your history, request get_recent_history to briefly review your latest browsing activity, or call get_most_visited to identify sites you visit most often, which can then be used to tailor responses or generate summaries. Each tool returns structured results suitable for feeding into LLM prompts, summaries, or personalized profiles.
How to install
Prerequisites:
- A computer with macOS, Windows, or Linux
- Python 3.8+ and pip
- uv (Rust-based environment) installed
- Access to Claude Desktop or your LLM chat app environment
Install and setup steps:
- Install uv and set up a virtual environment
# Install uv (if not installed) and create a venv uv venv uv pip install -r pyproject.toml uv pip install cryptography source .venv/bin/activate - Install the MCP server package (claude-desktop is used in the Claudite / Claude Desktop integration)
fastmcp install claude-desktop mcp_server.py - Restart Claude Desktop or your chat application to load the MCP server
# Depending on your setup, restart the host app or environment - Run the MCP server via uvx (if not automatically started)
uvx claude-desktop
Notes:
- Ensure your browser history databases are accessible in read-only mode as described in the privacy notes.
- You may need to configure paths to browser data if the automatic discovery doesn’t locate history files.
Additional notes
Tips and considerations:
- Privacy: The server reads local browser history in read-only mode and creates temporary copies to avoid locking issues. No data is transmitted externally.
- Browser support and paths: Browser data locations vary by OS and browser. You may need to manually configure file paths for Brave, Chrome, Edge, Opera, Arc, DuckDuckGo, Firefox, and Safari (macOS-specific).
- Platform coverage: The server supports Brave, Chrome, Edge, Opera, Arc, and DuckDuckGo across Windows, macOS, and Linux; Firefox is supported on all platforms; Safari support is macOS-only.
- If you see issues with histories not loading, check file permissions and ensure browsers aren’t running with locked history databases. Restart the browsers if needed.
- Environment variables: You can expose optional environment settings for local testing (e.g., HISTORY_PATH, DEBUG) if you customize the server.
Related MCP Servers
web-eval-agent
An MCP server that autonomously evaluates web applications.
mcp-neo4j
Neo4j Labs Model Context Protocol servers
Gitingest
mcp server for gitingest
zotero
Model Context Protocol (MCP) server for the Zotero API, in Python
fhir
FHIR MCP Server – helping you expose any FHIR Server or API as a MCP Server.
unitree-go2
The Unitree Go2 MCP Server is a server built on the MCP that enables users to control the Unitree Go2 robot using natural language commands interpreted by a LLM.