Quickstart This guide proves Termyte’s dependable local workflow before you try the
experimental runtime.
Prerequisites
Node.js 20 or later
npm
a repository where you can create .termyte/ local state
1. Install Termyte Confirm the CLI and local environment:
termyte --help termyte doctor
Missing optional agents or shells may appear as warnings. A doctor failure in
the shell runtime means you should not rely on experimental interception yet.
2. Check commands without executing them Run these inside a repository:
termyte check "cat .env" termyte check "git push --force origin main" termyte check "npm publish" termyte check "npm test"
Expected behavior:
Command Typical decision cat .envblock git push --force origin mainblock npm publishwarn npm testallow
check never executes the supplied command. It writes a redacted decision
event to .termyte/logs.jsonl.3. Add repository policy Preview a policy rule:
termyte policy local add "Ask before touching auth or payments" --dry-run
Save it:
termyte policy local add "Ask before touching auth or payments" --yes
This creates or updates termyte.policy.yaml in the current repository.
Review and test the effective policy:
termyte policy show termyte policy test "cat .env"
4. Add explicit memory Mark an otherwise allowed command as unsafe:
termyte mark-unsafe "npm test" termyte check "npm test" termyte memory
The later check warns because it exactly matches a command you marked unsafe.
Memory never weakens a policy block.
5. Review local decisions termyte logs termyte logs --blocked termyte logs --warned
6. Prepare a coding agent See the launch plan first:
termyte run --dry-run codex
Then launch a supported agent:
termyte run codex termyte run claude termyte run claudecode termyte run aider
Agent launch currently uses limited mode . Termyte prepares repository
policy, logs, memory, and session context, but full subprocess interception is
not guaranteed.
Next steps
Configure policies Set built-in, global, repository, and runtime policy.
Set up coding agents Understand supported agents, launch behavior, and limits.
Try governed execution Evaluate the experimental command and shell runtime.
Troubleshoot Diagnose installation, policy, and runtime problems.