Wiki Article

OpenAI Codex (AI agent)

Nguồn dữ liệu từ Wikipedia, hiển thị bởi DefZone.Net

Screenshot of the OpenAI Codex desktop application. A sidebar on the left lists threads, automations, and skills. The main panel shows a new-project start screen with suggested tasks and a text box for asking Codex to help build or summarise a project.
The OpenAI Codex desktop app start screen, showing options for creating a new project and using Codex tools

Codex is an AI coding agent developed by OpenAI and integrated with ChatGPT. Launched as a research preview in May 2025, it is designed to handle software engineering tasks – such as writing features, fixing bugs, and reviewing codebases – autonomously in a cloud environment before returning results for human review. The agent is available through ChatGPT's web app, a command-line interface, a desktop app for Windows and macOS, and several IDE integrations. In March 2026, OpenAI introduced Codex Security, an application-security agent designed to identify and propose fixes for software vulnerabilities.

Press coverage described Codex as OpenAI's entry into an increasingly competitive market for AI coding tools, where it competes with products such as Anthropic's similar Claude Code and Cursor. By March 2026, OpenAI reported that Codex had grown to more than 2 million weekly active users and was positioning it as a broader enterprise agent platform that could eventually be used for tasks beyond software development.

History

[edit]

OpenAI announced a research preview of Codex on May 16, 2025.[1] At launch it was powered by codex-1, a version of the o3 reasoning model optimized for software engineering, and could write features, answer questions about a codebase, fix bugs, and propose code changes for review.[2][3] The research preview first rolled out to some paid ChatGPT tiers, and ChatGPT Plus users gained access in June 2025.[2][4] In February 2026, OpenAI released a desktop Codex app that was intended to help users manage multiple coding agents over longer periods and use code to gather or analyze information.[5] GPT-5.3-Codex followed on February 5, 2026.[6] A week later it introduced GPT-5.3-Codex-Spark, a lower-latency variant for real-time interactive coding, initially available as a research preview for ChatGPT Pro users. InfoQ reported that Spark was OpenAI's first production model deployed on Cerebras hardware and ran about 15 times faster than earlier Codex versions.[7] GPT-5.4 followed in Codex on March 5.[8] Later in March 2026, Reuters reported that OpenAI planned to combine ChatGPT, Codex, and its browser into a single desktop "superapp" as part of an effort to simplify its product lineup and respond to growing competition from Anthropic.[9]

Features

[edit]

Each Codex task runs in a separate cloud environment preloaded with the user's repository, where the agent can read and edit files, run tests, and invoke other code-checking tools.[2] The design is intended to let Codex function more like a coding agent than a conventional assistant, handling tasks independently before returning results for review.[3] Most tasks take between 1 and 30 minutes; Codex returns command logs and test results so users can inspect what it has done.[2] Codex was initially available inside ChatGPT's web app and deliberately lacked general internet access for security reasons,[3] thoguht optional internet access was later enabled during task execution.[2]

By February 2026, OpenAI had described a single "App Server" architecture for Codex that powered the CLI, the VS Code extension, the web app, the macOS desktop app, and third-party IDE integrations including JetBrains and Xcode. This design was intended to keep long-running sessions and approval requests consistent across different client interfaces.[10] That same month, Apple added direct support for Codex in Xcode 26.3, allowing developers to use the agent inside Apple's development environment rather than only through OpenAI's own tools. TechCrunch reported that the integration let Codex inspect a project's structure, consult current Apple developer documentation, build projects, run tests, and revert changes through automatically created milestones.[11] Also in February, GitHub added Codex in public preview to its Agent HQ system, making the coding agent available inside GitHub, GitHub Mobile, and Visual Studio Code for some Copilot subscribers. The Verge reported that developers could assign Codex to issues and pull requests and compare its output with other agents such as Claude and Copilot.[12] Also that month, Figma announced an integration that let users move between Figma and Codex through Figma's Model Context Protocol server, connecting interface design and code implementation.[13]

In March 2026, SecurityWeek reported that Codex Security had been in private beta since 2025, including testing by Netgear, and that it worked by first building a threat model of a repository before looking for vulnerabilities and proposing fixes. OpenAI said it had tested the tool on 1.2 million commits over the previous 30 days, identifying nearly 800 critical vulnerabilities and more than 10,000 high-severity issues in projects including Chromium, OpenSSL, PHP, the self-hosted Git service GOGS, and GnuTLS.[14] Help Net Security reported that Codex Security validates suspected issues in sandboxed environments so it can prioritize findings by likely real-world impact and reduce false positives before suggesting patches.[15] Axios reported that the tool was derived from an earlier internal project called Aardvark.[16]

Reception

[edit]

Reuters described coding as one of artificial intelligence's most commercially successful uses and said the market for coding agents had become a key battleground among AI companies.[5][4][17] OpenAI's desktop Codex app, released in February 2026, was part of the company's effort to gain ground against Anthropic in AI coding tools. By that point, more than one million developers had used Codex in the previous month as competition intensified with tools such as Claude Code and Cursor, which together had become a popular way for engineers to quickly create new services.[5][18][19] In March 2026, WIRED described Codex as part of OpenAI's effort to catch up with Anthropic, and reported that the company hoped eventually to use Codex not only for programming but also for task-completing features across its products.[20] That same month, Fortune reported that OpenAI said Codex had grown to more than 1.6 million weekly active users after the release of GPT-5.3-Codex, and that companies including Cisco, Nvidia, Ramp, Rakuten, and Harvey had deployed it across developer teams. Fortune added that OpenAI was presenting Codex as a broader enterprise agent platform that could eventually be used for tasks beyond software development.[21] Reuters reported that month that OpenAI had agreed to acquire Python toolmaker Astral, a deal that would bring Astral's developer tools into Codex. OpenAI said Codex had by then surpassed 2 million weekly active users, with usage up fivefold since the beginning of 2026.[22]

References

[edit]
  1. ^ Knight, Will (May 16, 2025). "OpenAI Launches an Agentic, Web-Based Coding Tool". Wired. Retrieved March 16, 2026.
  2. ^ a b c d e "Introducing Codex". OpenAI. May 16, 2025. Retrieved March 12, 2026.
  3. ^ a b c Heath, Alex (May 16, 2025). "ChatGPT is getting an AI coding agent". The Verge. Retrieved March 12, 2026.
  4. ^ a b Babu, Juby (August 19, 2025). "Musk's xAI forays into agentic coding with new model". Reuters. Retrieved March 12, 2026.
  5. ^ a b c Seetharaman, Deepa (February 2, 2026). "OpenAI launches Codex app to gain ground in AI coding race". Reuters. Retrieved March 12, 2026.
  6. ^ Axon, Samuel (February 5, 2026). "With GPT-5.3-Codex, OpenAI pitches Codex for more than just writing code". Ars Technica. Retrieved March 6, 2026.
  7. ^ De Simone, Sergio (March 2, 2026). "OpenAI Codex-Spark Achieves Ultra-Fast Coding Speeds on Cerebras Hardware". InfoQ. Retrieved March 21, 2026.
  8. ^ "Introducing GPT-5.4". OpenAI. May 3, 2026. Retrieved March 16, 2026.
  9. ^ "OpenAI plans desktop 'superapp' to streamline user experience". Reuters. March 19, 2026. Retrieved March 21, 2026.
  10. ^ Stiller, Eran (February 17, 2026). "OpenAI Publishes Codex App Server Architecture for Unifying AI Agent Surfaces". InfoQ. Retrieved March 12, 2026.
  11. ^ Perez, Sarah (February 3, 2026). "Xcode moves into agentic coding with deeper OpenAI and Anthropic integrations". TechCrunch. Retrieved March 16, 2026.
  12. ^ Warren, Tom (February 4, 2026). "GitHub adds Claude and Codex AI coding agents". The Verge. Retrieved March 20, 2026.
  13. ^ Mehta, Ivan (February 2, 2026). "Figma partners with OpenAI to bake in support for Codex". TechCrunch. Retrieved March 12, 2026.
  14. ^ Kovacs, Eduard (March 10, 2026). "OpenAI Rolls Out Codex Security Vulnerability Scanner". SecurityWeek. Retrieved March 16, 2026.
  15. ^ Markovic, Sinisa (March 9, 2026). "OpenAI joins the race in AI-assisted code security". Help Net Security. Retrieved March 21, 2026.
  16. ^ Sabin, Sam (March 16, 2026). "OpenAI unveils Codex Security to automate code security reviews". Axios. Retrieved March 12, 2026.
  17. ^ Nellis, Stephen (May 19, 2025). "Microsoft to offer rival AI models from own data center; launches AI coding agent". Reuters. Retrieved March 12, 2026.
  18. ^ "OpenAI CEO Altman dismisses Moltbook as likely fad, backs the tech behind it". Reuters. February 3, 2026. Retrieved March 12, 2026.
  19. ^ Bensinger, Greg (November 25, 2025). "Amazon pushes in-house AI coding tool Kiro over competitors', memo shows". Reuters. Retrieved March 12, 2026.
  20. ^ Zeff, Maxwell (March 11, 2026). "Inside OpenAI's Race to Catch Up to Claude Code". WIRED. Retrieved March 12, 2026.
  21. ^ Kahn, Jeremy (March 4, 2026). "OpenAI sees Codex users spike to 1.6 million, positions coding tool as gateway to AI agents for business". Fortune. Retrieved March 12, 2026.
  22. ^ Singh, Jaspreet (March 19, 2026). "OpenAI to buy Python toolmaker Astral to take on Anthropic". Reuters. Retrieved March 20, 2026.