MCP-PROCESS

MCP-PROCESS is an MCP server that provides shell and process access to LLM clients such as Claude Desktop. It connects an AI agent directly to local command-line environments like Windows Subsystem for Linux or utilities like PostgreSQL CLI, enabling it to execute terminal commands, run filesystem utilities, and return command output. Developers, system administrators, and technical users use it to automate local development tasks, inspect directory structures, perform search operations, and manipulate files through natural language queries. The server allows fine-grained configuration of the target process binary, execution timeouts, custom tool names, output regex filters, and a list of forbidden words to prevent specific commands from running. Because it gives the model direct control over a local shell session and can modify or delete critical system files, it is primarily intended for controlled local environments and testing rather than untrusted production workflows.

Category: Cloud & Infrastructure

Tags: automation, bash, filesystem, shell, terminal

Visit MCP-PROCESS

How to install and configure MCP-PROCESS

To set up and run MCP-PROCESS with Claude Desktop: 1. Ensure Python 3.10 or higher is installed, along with WSL on Windows (pywinpty required) or ptyprocess on Linux/macOS. 2. Clone the repository and install the package: bash git clone https://github.com/jrame/mcp-process.git cd mcp-process pip install . 3. Open your Claude Desktop configuration file (e.g., %AppData%/Claude/claude_desktop_config.json). 4. Add mcp-process under the mcpServers object: json { "mcpServers": { "wsl": { "command": "mcp-process", "args": [ "--process-path-args", "wsl.exe --cd /mnt/c/Users/YourName", "--filter-patterns", "\\x1b\\[[0-9;]*m", "--exec-name", "exec", "--exec-description", "Exécute une commande statique", "--exec-timeout", "60" ] } } } 5. Restart Claude Desktop to apply the configuration.

What you can do with MCP-PROCESS

• Running terminal directory commands such as ls, pwd, and tree to navigate and inspect local project structures. - Executing search and inspection tools like grep, find, and ripgrep to scan codebase contents directly through Claude. - Interacting with command-line database clients like psql to execute targeted SQL queries and review output tables. - Running build and version control tools like git directly within WSL sessions to check repository statuses.

Key facts

• https://github.com/jrame/mcp-process
• Cloud & Infrastructure, Files, Documents & PDFs
• automation, bash, filesystem, shell, terminal

Part of MCP Servers

How do I install MCP-PROCESS?

Clone the repository using git, navigate into the project directory, and install it with pip install dot. Ensure you have Python 3.10 or higher and either pywinpty on Windows or ptyprocess on Linux or macOS.

What can MCP-PROCESS do?

It provides shell execution access to an MCP client. This allows the AI model to run terminal commands, execute file system operations, pass arguments to tools like WSL or psql, and filter ANSI escape sequences from terminal outputs.

Which MCP clients work with MCP-PROCESS?

MCP-PROCESS is designed and tested for use with Claude Desktop via the claude desktop config json file, but any Model Context Protocol compliant client that supports local command-based server configurations can integrate with it.

Is MCP-PROCESS open source?

Yes, MCP-PROCESS is open source software released under the MIT license. You can inspect its source code, contribute fixes, or modify it on its official GitHub repository.

What security precautions should I take with MCP-PROCESS?

Because the server grants direct shell and filesystem execution permissions, the author recommends running it only in controlled or isolated environments. Built-in mechanisms include timeouts, regex filters, and a forbidden words list, though full isolation like WSL is advised.