> ## Documentation Index
> Fetch the complete documentation index at: https://docs.obversa.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Codex Engine

> @obversa/engine-codex runs engine requests through a fresh Codex CLI process.

`@obversa/engine-codex` is an engine plugin that runs requests through the `codex` command-line interface. It provides an OpenAI-backed engine for verification and review seats.

## Requirements

* Node.js 22.12 or later
* A local Codex CLI installation with existing authentication

## Install

```bash theme={null}
pnpm add @obversa/engine-codex
```

## Identity

The plugin serves the OpenAI provider. It reports `provider: 'openai'` and uses `request.model` or `options.defaultModel`. When neither is set, the CLI runs without an explicit model flag.

## Usage

```ts theme={null}
import { CodexEngine } from '@obversa/engine-codex';

const engine = new CodexEngine({
  defaultModel: 'gpt-5.4',
});
```

Pass the engine instance into the Obversa runtime.

## Limits

* The engine runs in a read-only sandbox by default; write permissions require `permissionMode: 'bypassPermissions'`.
* Subprocess execution requires a working Codex binary in `PATH` or an explicit `cliBinary` path.
* Large prompts pass through standard input, and final output is read from a temporary file.