One command turns any folder into an MCP server. Give Claude/Cursor or Custom agent, and any AI client real access to your files, terminal, and system.
Install globally, point it at a folder, paste the URL into your AI. That's it.
One npm command installs the openlol CLI globally on your machine. Works on macOS, Linux, and Windows.
Point it at any folder. openlol starts a local MCP server with a unique session token for secure access.
Paste the MCP URL into Claude, Cursor, or any MCP-compatible AI client. The AI now has access to your project.
Every tool is scoped to the folder you chose — the AI can't escape the sandbox.
Run any shell command in the scoped working directory. Build, test, lint, deploy — whatever you need.
Read the full text content of any file in the project folder.
Create or overwrite files. Auto-creates parent directories as needed.
Append content to the end of an existing file or create it if missing.
Delete files or directories recursively within the scoped folder.
List all files and subdirectories at any path in the project.
Create directories with automatic parent creation (mkdir -p style).
Move or rename files and directories within the project scope.
Check whether a file or directory exists before reading or writing.
OS, architecture, memory usage, uptime, and Node.js version at a glance.
Disk storage usage for all mounted drives and partitions.
List env vars with optional substring filtering. Debug configs instantly.
openlol bridges the gap between AI chat windows and your actual dev environment.
Point openlol at a project folder and ask Claude to review your code. It reads files, understands structure, and gives context-aware feedback — no copy-pasting needed.
Describe the project you want and let the AI build the entire file structure, configs, and boilerplate. It creates directories, writes files, and runs setup commands.
Ask the AI to check your system health — disk space, memory, running processes, env vars. Useful for debugging deployment issues on remote machines.
Point it at your logs directory and ask the AI to find errors, patterns, or anomalies. It can read, grep, and parse log files to pinpoint issues.
Let the AI read your codebase, understand patterns, and apply changes across multiple files. Rename variables, update imports, migrate APIs — hands-free.
Ask the AI to run your test suite, read the output, fix failing tests, and re-run until everything passes. Full loop, no context switching.
openlol runs a local HTTP server that speaks the Model Context Protocol.
Works with Claude, Cursor, and any MCP-compatible client.
Every session gets a unique UUIDv4 token. Tools are sandboxed to the folder you choose.
No setup files, no environment variables, no cloud accounts. Install and run.
Streamable HTTP (modern) and SSE (legacy) endpoints. Works with every MCP client.
macOS, Linux, Windows. Node.js 18+ is the only requirement.
Install openlol and give your AI client real access to your projects.