AiMatch

Termaxa

Termaxa previews, backs up, and gates shell commands and supported file writes from AI coding agents, then records the decisions. Developers using Claude Code, Codex, Cursor, or Copilot can inspect destructive actions before they run.

One of 140 tools in Developer Tools

Visit Termaxa Compare alternatives Own this tool? Claim it
Termaxa screenshot

Who it's for

  • Developers using Claude Code who want command consequences previewed and covered file changes backed up before approval.
  • Teams using Cursor who need policy checks for shell events and supported native Write, Edit, or Delete operations.
  • Developers using Codex who need hook-based denial and audit records, and can work without an ask verdict because Codex treats ask as refusal.

Not the right fit if…

  • Codex teams that require an interactive ask verdict from the hook: the page says Codex hooks can only deny, so an ask is a refusal.
  • Teams that need shared policies and approvals through Termaxa Cloud today: the page lists this cloud offering as planned.

How it fits your workflow

Install Termaxa and run termaxa init in a project to write a policy and hook for a detected harness. Run termaxa doctor to verify the wiring, or termaxa replay to judge commands from existing agent transcripts.

When an agent proposes a covered shell command or native file write, Termaxa inspects the action, previews its consequences, applies the policy, and backs up covered files before approval. It can deny an action and give the reason to the agent.

Termaxa records decisions outside the repository. Users can review those records, use rollback for covered changes, and inspect execution reports.

Pricing

Free: The core tool is free forever and open source under MIT/Apache-2.0. The page says Termaxa Cloud for shared policies and approvals is planned, not currently available.

Termaxa

Free forever

  • · Open source under MIT/Apache-2.0
  • · No feature gates or license keys

Prices checked on Oct 8, 2026 from the vendor's site. They can change; confirm before you buy.

Key features

Command consequence previews
The termaxa check command evaluates a command without executing it and can show the files or other consequences involved. Examples on the page include destructive file commands, force pushes, and database operations.
Backups and rollback
Termaxa backs up affected files before an approved destructive action and supports rollback. The page also describes database dump and rollback handling.
Policy-based decisions
Policies can allow, ask about, or deny actions. The page describes a closed-side option for unattended runs, with unrecognized actions and backup failures set to deny.
AI coding-agent hooks
Hooks are available for Claude Code, Codex, Cursor, and Copilot. The page notes that Codex hooks can deny but cannot provide an ask decision, while Copilot can present an ask as a prompt.
Native file-tool inspection
Termaxa can judge supported native Write, Edit, and Delete operations and patch file headers using path rules, with backups and records for covered writes.
Compound-command and intent inspection
Compound shell commands are split and judged by segment. The page also describes classifying intent across alternate spellings and using a per-session circuit breaker when an agent repeats a blocked action.

Works with

  • Claude Code
  • Codex
  • Cursor
  • GitHub Copilot
  • Herdr

Limitations to know

  • Codex hooks can only deny; an ask decision is a refusal in Codex.
  • Codex uses its login shell by absolute path and does not offer the shell-setting lever described for Claude Code, so the page says Codex is gated through hooks instead.
  • Termaxa Cloud for shared policies and approvals is listed as planned rather than available.
  • The page says a file no rule names is silence by design.

Alternatives to consider

  • Execlave

    Choose it if you need a broader AI-agent governance platform with managed cloud or self-hosted deployment, rather than Termaxa's command-focused previews, backups, and shell hooks.

  • Kastra

    Choose it if you need authorization decisions for agents, models, and AI tools before they act, rather than a gate focused on coding-agent shell commands and file-write consequences.

Compare Termaxa side by side

Getting started

brew install termaxa/tap/termaxa
  1. Install Termaxa with a supported package manager; the page provides a Homebrew command and lists prebuilt binaries for Linux, macOS, and Windows.
  2. Run termaxa init in a project to create the policy and hook for a detected harness.
  3. Run termaxa doctor to check that the gate is wired up.
  4. Run termaxa replay to judge commands from existing agent transcripts before committing to a policy.

Frequently asked questions

Does Termaxa execute a command when I use termaxa check?

No. The page says termaxa check reads the command and answers; nothing executes.

Which AI coding agents are named as supported?

The page names Claude Code, Codex, Cursor, and Copilot.

Is Termaxa free?

The page says the tool is free forever, open source under MIT/Apache-2.0, and has no feature gates or license keys.

Can Codex use Termaxa ask decisions?

No. The page states that Codex hooks can only deny, so an ask is a refusal there.

Is Termaxa Cloud available?

The page lists Termaxa Cloud for shared policies and approvals as planned, with the item unchecked.