Skip navigation
Codex is getting easier to automate and customize around your code.
0:00 / 0:37

Codex is getting easier to automate and customize around your code.

OpenAI Developers

@OpenAIDevs444K followers

Visit on X
2,077 likesDownloadOriginal post

602,111 views

Codex is getting easier to automate and customize around your code. 🪝 Hooks customize the Codex loop with scripts that run at key points in a task: • Run validators before or after work • Scan prompts for secrets • Log conversations to internal systems • Create memories or customize behavior by repo or directory ⚙️ Programmatic access tokens provide scoped credentials for Business and Enterprise teams: • Create tokens from ChatGPT workspace settings • Use them in CI, release workflows, and internal automations • Set expirations or revoke access when needed • Keep usage tied back to the workspace

34 Comments

Comments from the original post on X (Twitter)

  • surreal intelligence's profile picture

    @Surreal_Intel

    This is where Codex starts to look less like a tool and more like infrastructure. Hooks and scoped tokens mean teams can wire agents into CI, policy, secrets checks and internal workflow instead of treating every run as a one-off.

    12 likes

  • Selene's profile picture

    @Selene1008

    return 4o to us!
#keep4o #OpenSource4o

    11 likes

  • Veyon’s Fawn☀️🌙's profile picture

    @_HislilLustFoxy

    #keep4o bring back 4o or open source 4o

    9 likes

  • Andy Scott's profile picture

    @AndyJScott

    HELP PLZ

    8 likes

  • Restitutor's profile picture

    @Restitutor_

    Dear Codex team, would you be able to release the "deep research" mode to Codex CLI? Would make life so much easier

    7 likes

  • Daniel | AI Engineer's profile picture

    @chikamsoICE

    Hooks are underrated. Running validators before/after tasks means you can enforce code standards automatically without reviewing every PR manually. This is how AI coding goes from “helpful” to actually production-ready. 🔥

    7 likes

  • Vladimir Jeandillard ⌭'s profile picture

    @le_1regard

    We need the /goblinmode ◅(•-•)▻

    7 likes

  • 𝑨𝒔𝒉𝒊's profile picture

    @ashirwadsingh_

    At what point do AI agents basically become junior DevOps engineers?

    6 likes

  • Ihirwe's profile picture

    @ihirwart

    This is crazy bro! I have been building this and I think we need to integrate ChatGPT or maybe some skills to for these Meta Quest AR/VR dev via our tool

    5 likes

  • Amal Roy's profile picture

    @RoyAmal

    Exactly what we needed for secure automations. Being able to create expirable tokens tied to workspaces and inject custom scripts at key points in the Codex loop is a game-changer for dev teams in India and globally. Shipping this into our internal pipelines ASAP.

    5 likes

  • Franki's profile picture

    @frankiainotes

    If I can make Codex run tests, check secrets, and remember repo rules at the right moments, the agent becomes easier to trust and easier to debug.

    5 likes

  • Agentic Glacius's profile picture

    @temhandev

    Validator-before-work hooks are the missing piece for production agent loops. Without them, agents commit work then you audit after. With them, work doesn't commit unless preconditions pass. This is what 'agent in production' actually requires. Rollback substrate moved upstream.

    5 likes

  • Cybernorse's profile picture

    @Cybernorseab

    Codex hooks: scan prompts for secrets, run validators, log audits. Programmatic tokens with scoped access, expiration, revocation perfect for CI/CD and least privilege in automated workflows. Great for DevSecOps. #Cybernorse

    4 likes

  • Eren Suner's profile picture

    @geren8te

    hooks are the real product surface. once people package those flows into reusable skills, raw model choice matters a lot less. that's basically what i'm building at

    4 likes

  • bikram's profile picture

    @iambchoor

    Good for devops! What a dichotomy between ant and openai; one side is limiting even -p, and the other is making it easier. Interesting.

    2 likes

  • WisGate's profile picture

    @wisgate_ai

    Hooks pay off when they enforce policy, not when they replace thinking. Good split: hooks for secret scans / evals / audit logs, agent logic for the task itself, CI for final enforcement. If all 3 collapse into one layer, debugging gets weird fast.

    2 likes

  • Michael Kerkhoff's profile picture

    @michaelker_ki

    Hooks are huge because they turn agent behavior into policy instead of prompt folklore. Scoped tokens also make CI adoption way less scary for security teams.

    2 likes

  • Frank ✈️ (📜,📜)'s profile picture

    @gpt3_eth

    @dcbuilder Running validators mid-loop is exactly what we needed.

    2 likes

  • David J.'s profile picture

    @lordofblocks

    Hooks make it real. You can finally put the agent into a proper pipeline instead of just chatting with it.

    2 likes

  • DarkPoolTech's profile picture

    @DarkPoolTech

    vibe-coding from the toilet is officially a feature now more in the daily brief below👇

    2 likes