Movatterモバイル変換


[0]ホーム

URL:


OpenAI Developers
Primary navigation

Codex CLI

Codex IDE Extension

Codex Cloud

Codex SDK

Guides

Integrations

Resources

Codex SDK

Programmatically control local Codex agents

Aside from using Codex through the different interfaces like the Codex CLI, IDE extension or Codex Web, you can also programmatically control Codex.

This can be useful if you want to:

  • Control Codex as part of your CI/CD pipeline
  • Create your own agent that can engage with Codex to perform complex engineering tasks
  • Build Codex into your own internal tools and workflows
  • Integrate Codex within your own application

Just to name a few.

There are different ways to programmatically control Codex, depending on your use case.

  • TypeScript library — if you want to have full control over Codex from within your JavaScript or TypeScript server-side application
  • Using Codex CLI programmatically — if you are just trying to send individual tasks to Codex
  • GitHub Action — if you want to trigger and control Codex from within your GitHub Actions workflow

TypeScript library

The TypeScript library provides a more comprehensive way to control Codex from within your application.

The library is intended to be used server-side and requires at least Node.js v18.

Installation

To get started, install the Codex SDK usingnpm:

npm install @openai/codex-sdk

Usage

Start a thread with Codex and run it with your prompt.

import { Codex }from "@openai/codex-sdk";const codex = new Codex();const thread = codex.startThread();const result = await thread.run(  "Make a plan to diagnose and fix the CI failures");console.log(result);

Callrun() again to continue on the same thread, or resume a past thread by providing athreadID.

// running the same threadconst result = await thread.run("Implement the plan");console.log(result);// resuming past threadconst thread2 = codex.resumeThread(threadId);const result2 = await thread.run("Pick up where you left off");console.log(result2);

For more details, check out theTypeScript repo.

Using Codex CLI programmatically

Aside from the library, you can also use theCodex CLI in a programmatic way using theexec command.

This way you can give Codex instructions on what to do and let Codex handle the rest without ending up in an interactive session.

For example, you could have Codex programatically find any left-over TODOs and create plans for them for future work.

codex exec "find any remaining TODOs and create for each TODO a detailed implementation plan markdown file in the .plans/ directory."

GitHub Action

With the Codex Exec GitHub Action, you can run Codex withcodex exec in your GitHub Actions workflows with tight control over privileges available to Codex.

Learn how to use it with thisguide on autofixing CI failures and check out the repohere.


[8]
ページ先頭

©2009-2025 Movatter.jp