Supported agents
Belay 0.7.0 supports four AI coding agents. Run belay setup in your project and Belay configures each one it detects automatically.
Belay integrates with Claude Code's tool hooks, intercepting Bash commands and file edits before they execute.
Belay guards Cursor's shell execution. Commands are evaluated before Cursor runs them.
Belay integrates with VS Code's agent tool hooks to guard commands issued through GitHub Copilot.
Belay guards Codex shell invocations before they run.
How to use Belay with Claude Code
Run belay setup in your project directory. Belay detects your Claude Code configuration and wires the integration automatically. You do not need to edit any configuration file by hand.
$ cd your-project $ belay setup
How to use Belay with Cursor
The same belay setup command handles Cursor. It detects your Cursor configuration and adds the Belay integration without overwriting your existing settings.
How to use Belay with GitHub Copilot (VS Code)
Run belay setup in your project. Belay detects the VS Code agent configuration and wires itself in. The integration works with GitHub Copilot's agent tool execution.
How to use Belay with Codex
Run belay setup and Belay auto-detects your Codex configuration. All Codex shell commands pass through Belay before they run.
Targeting specific agents
If you want to configure only specific agents rather than all detected ones:
$ belay setup --surfaces claude,cursor
Valid surface names are: claude, cursor, vscode, codex.
The same decision for every agent
Regardless of which agent you use, Belay applies the same evaluation to every command. The decision — allow, confirm, hold, or deny — does not depend on which agent issued the command. Your policy applies consistently across all surfaces.