Cursor History MCP

MCP server for browsing, searching, exporting, and backing up your Cursor AI chat history.

Bring your Cursor AI chat history directly into Claude. Search past conversations, export sessions, create backups, and generate year-in-review reports—all through natural language. Built on the Model Context Protocol for seamless AI assistant integration.

Free, open-source, and MIT licensed. Built by the community, for the community.

# Why This Project?

There are other Cursor history tools out there (like the Python-based Cursor-history-MCP). Here's what makes this one different:

| Feature | cursor-history-mcp (this project) | Other Solutions |
|- -- -- -- --|- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --|- -- -- -- -- -- -- -- --|
| Setup | npx cursor-history-mcp - zero install | Docker, Python, dependencies |
| Speed | Instant - direct SQLite reads | Slow - requires LLM vectorization |
| Search | Grep-style text matching - precise & stable | Vector retrieval - unpredictable results |
| LLM Required | No - works offline | Yes - needs Ollama/embeddings |
| Language | TypeScript (type-safe) | Python |
| Backup/Restore | Built-in | Not available |
| Migration | Move sessions between workspaces | Not available |
| Dependencies | Minimal (just Node.js) | Docker, LanceDB, Ollama, FastAPI |

# # Key Advantages

• Blazing Fast: No embedding or vectorization step. Reads directly from Cursor's native SQLite database, so results are instant.
• Grep-Style Search: Uses direct text matching instead of vector retrieval. More lightweight, predictable, and stable for most use cases—no hallucinated results, no embedding drift, and exact matches every time.
• Zero Configuration: Run with npx - no Docker containers, no Python environments, no API keys, no LLM setup.
• Works Offline: Everything runs locally without any external services or AI models.
• Data Portability: Full backup, restore, and cross-workspace migration capabilities to keep your chat history safe and portable.
• Lightweight: ~50KB package vs multi-GB Docker images with vector databases.

# Installation

No installation required! Run directly via npx:

npx cursor-history-mcp

# Configuration

# # Cursor

# # Claude Code

Add to your Claude Code MCP settings:

{
  "mcpServers": {
    "cursor-history": {
      "command": "npx",
      "args": ["-y", "cursor-history-mcp"]
    }
  }
}

# # Claude Desktop

Add to your Claude Desktop configuration (~/.claude/claude_desktop_config.json):

{
  "mcpServers": {
    "cursor-history": {
      "command": "npx",
      "args": ["-y", "cursor-history-mcp"]
    }
  }
}

# Available Tools

| Tool | Description |
|- -- -- -|- -- -- -- -- -- --|
| cursor_history_list | List chat sessions with metadata |
| cursor_history_show | View full conversation content |
| cursor_history_search | Search across all sessions |
| cursor_history_export | Export session to Markdown or JSON |
| cursor_history_backup | Create backup of all history |
| cursor_history_restore | Restore from backup (destructive) |
| cursor_history_migrate | Move/copy sessions between workspaces (destructive) |
| cursor_history_year_pack | Generate year-in-review data package with stats, topics, and prompt template |

# Usage Examples

After configuring, ask your AI assistant:

• "List my Cursor chat sessions"
• "Show me session # 1"
• "Search my Cursor history for 'authentication'"
• "Export session # 1 as markdown"
• "Backup my Cursor chat history"

# Requirements

• Node.js 20+
• Cursor IDE installed with existing chat history

# Contributing

Contributions are welcome! Whether it's bug reports, feature requests, documentation improvements, or code contributions—all PRs and issues are appreciated.

• Open an issue
• Submit a pull request

# License

MIT