askme-cli

ASKME-CLI is an MCP server that provides a user confirmation interface directly from the terminal. When integrated with an MCP client, the AI assistant can request a user's confirmation or next steps by invoking the ask_me tool. The tool opens a dedicated terminal-based confirmation UI, allowing the user to input their next plan or confirmation and submit it back to the assistant. This enables a smooth, interactive loop where the AI can pause for human input before proceeding. You can customize the terminal experience by setting the ASKME_CLI_TERMINAL environment variable to your preferred terminal (Warp, iTerm2, Terminal, Kitty, Alacritty, Hyper, etc.). The server supports receiving text and image inputs via the ask_me tool, making it suitable for confirming tasks, plans, or decisions that require human validation.

Prerequisites:

• Node.js and npm installed on your machine
• Access to the MCP client configuration workflow

Step-by-step:

1. Install the package from npm (or clone the repository if you have the source): npm install -g askme-cli

2. Build the project if a build step is provided by the repo: npm run build

3. Start the CLI server (development or production as appropriate): npm run start

4. Verify the server is reachable through MCP by following the configuration snippet in the README's Quick Setup section and adding the server entry to your MCP client config as shown: { "mcpServers": { "askme-cli": { "command": "npx", "args": ["askme-cli"], "env": { "ASKME_CLI_TERMINAL": "warp" } } } }

Note:

• Ensure your terminal application supports opening new windows/tanes per the chosen terminal setting (e.g., Warp, iTerm2).
• If you encounter Windows compatibility issues, check PRs or issues for compatibility fixes and adjust the env or command invocation accordingly.

Tips and notes:

• The server exposes a single MCP entry point named askme-cli. Use the provided MCP config example to route calls to the ask_me tool.
• Customize the terminal interface by setting ASKME_CLI_TERMINAL to your preferred terminal to ensure a smooth user experience.
• The ask_me tool accepts a what_we_did field (summary of AI work) and returns the user-provided next steps or confirmation. You can incorporate this response into your subsequent AI prompts or actions.
• If you need to adapt the command for different environments, you can swap npx with a local install or use a different launcher, but ensure the MCP config reflects the correct command and arguments.
• Monitor logs for troubleshooting: look for initialization messages, terminal spawning, and user input handling to diagnose common issues.