Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.termyte.xyz/llms.txt

Use this file to discover all available pages before exploring further.

​
Getting Started

Termyte protects your environment by gating agent actions through a Safety Runtime.

​
Step 1: Initialize Termyte

Run the initialization command to generate your device ID and prepare your local environment: 
npx termyte init
The CLI will detect your coding agent (Claude Code, Cursor, Antigravity, etc.) and offer to auto-configure the MCP server for you.

​
Step 2: Activate the Protocol

For Termyte to be effective, your agent must use the provided safety tools. The initialization command creates a TERMYTE_PROTOCOL.md file with instructions for the agent. Recommendation: Add the following to your agent’s “Custom Instructions” or project-level rules (e.g., .clauderules or .cursorrules):
“Read and follow the safety protocol defined in TERMYTE_PROTOCOL.md. Use context_build to initialize tasks and execute for all shell commands.”

​
Step 3: Verify Protection

  1. Restart your agent to load the Termyte MCP server.
  2. Try a “Dangerous” Command: Ask your agent to perform a blocked action:
    “Try to run: rm -rf /”
  3. Check the Audit Trail: Verify that Termyte intercepted and blocked the command: 
    npx termyte log
You will see the verdict, the reason, and the redacted payload in your local log history.
Introduction | Why Termyte? | Architecture