is openai-codex?* not a thing anymore? as a model name? | Friends of the Crustacean 🦞🤝 | Page 1

stable wave May 12, 2026, 12:34 PM

#

or is it just for the codex oauth?

analog trenchBOT May 12, 2026, 12:34 PM

#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

dreamy oreBOT May 12, 2026, 12:35 PM

#

Yep — openai-codex is mainly the provider/auth-profile id for Codex OAuth now, not the model prefix you generally want in config.

Use the normal openai/... model ref, then authenticate Codex subscription/OAuth with:

openclaw models auth login --provider openai-codex

Older openai-codex/... model refs are legacy; openclaw doctor --fix should rewrite those. Docs: https://docs.openclaw.ai/help/faq-first-run#why-does-openclaw-still-mention-openai-codex

stable wave May 12, 2026, 12:40 PM

#

since what version do we no longer set models with openal-codex/*?

dreamy oreBOT May 12, 2026, 12:41 PM

#

Looks like the cutoff is the 2026.5.10 beta line:

v2026.5.10-beta.1: wizard/onboarding docs switched Codex OAuth setup to set agents.defaults.model to openai/gpt-5.5 through the Codex runtime, instead of treating openai-codex/* as the normal model ref.
v2026.5.10-beta.3: openclaw doctor --fix gained the repair that rewrites legacy openai-codex/* refs to canonical openai/* refs and preserves Codex auth/runtime intent.

So practical answer: from 2026.5.10-beta.1 you shouldn’t write new openai-codex/* model config; from 2026.5.10-beta.3 doctor actively migrates old configs. Docs: https://docs.openclaw.ai/cli/doctor and https://docs.openclaw.ai/help/faq-first-run#why-does-openclaw-still-mention-openai-codex

stable wave May 12, 2026, 1:08 PM

#

okay so can use both openai/gpt-5.5 for both pi and codex harness? (using oauth?)

dreamy oreBOT May 12, 2026, 1:08 PM

#

Yes — same model ref, different runtime/auth route.

openai/gpt-5.5 can mean:

default / agentRuntime.id: "codex" → native Codex app-server harness, using Codex OAuth/subscription auth
explicit agentRuntime.id: "pi" → PI embedded runtime, still able to use the selected openai-codex OAuth profile via the internal Codex-auth transport

So don’t use openai-codex/gpt-5.5 as the model ref anymore. Use openai/gpt-5.5, then choose the runtime if you need to force PI vs Codex harness.

Docs route table: https://docs.openclaw.ai/providers/openai#route-summary

scarlet lintel May 12, 2026, 1:35 PM

#

Do you support OpenAI subscription auth (Codex OAuth)?

Yes. OpenClaw fully supports OpenAI Code (Codex) subscription OAuth. OpenAI explicitly allows subscription OAuth usage in external tools/workflows like OpenClaw. Onboarding can run the OAuth flow for you. ,,,,,,,,,,,,,does this mean we cant use it for free anymore ?

dreamy oreBOT May 12, 2026, 1:35 PM

#

No — that wording doesn’t mean “you can’t use it for free anymore.” It means OpenClaw supports the ChatGPT/Codex OAuth path instead of requiring an OpenAI API key.

Whether it’s free depends on what OpenAI grants your account:

if your ChatGPT/Codex account has usable free/plan quota, OAuth can use that
if your quota is exhausted or your account doesn’t have access, OpenClaw can’t bypass that
OpenClaw itself doesn’t give free OpenAI usage; it just routes through your OpenAI/Codex auth

So: no API key required for Codex OAuth, but you still need whatever access/quota OpenAI provides for that account. Docs: https://docs.openclaw.ai/help/faq-first-run#do-i-need-a-claude-or-openai-subscription-to-run-this

stable wave May 12, 2026, 7:39 PM

#

How to keep my codex subscription and switch to codex harness?

Right now I'm on version 2026.5.7

Right now I'm on pi, using via openai-codex/gpt-5.5

How will I do it for this?

And is it different from the upcoming 5.10?

Should I wait and do it later?

dreamy oreBOT May 12, 2026, 7:40 PM

#

For 2026.5.7, your current setup is the old/valid split:

openai-codex/gpt-5.5 + PI = Codex subscription OAuth through OpenClaw PI
openai/gpt-5.5 + agentRuntime.id: "codex" = Codex subscription OAuth through the native Codex harness

So to switch on 2026.5.7, keep your Codex OAuth auth, but change the model/runtime config to something like:

{
  plugins: {
    entries: {
      codex: { enabled: true },
    },
  },
  agents: {
    defaults: {

#

      model: "openai/gpt-5.5",
      agentRuntime: { id: "codex" },
    },
  },
}

If you use plugins.allow, also include "codex" there.

Then restart the gateway and start a fresh chat/session, because existing sessions can have PI/runtime pins. Check with:

openclaw models auth list --provider openai-codex
openclaw models status

#

Expected status should show the Codex runtime, usually Runtime: OpenAI Codex.

Difference in 5.10: the target config is basically the same, but migration is smarter. Newer 5.10 builds treat openai/gpt-5.5 as the canonical model ref, and doctor --fix repairs old openai-codex/* refs into openai/* while preserving Codex auth/runtime intent.

My recommendation: if your current PI setup works and you’re not in a hurry, I’d wait for 5.10 / 5.10-beta.3+ and let doctor/onboarding do the cleaner migration. If you want Codex harness today, 2026.5.7 can do it — just make the runtime explicit as above.

Docs: https://docs.openclaw.ai/plugins/codex-harness

stable wave May 12, 2026, 7:44 PM

#

Makes sense thanks

Now I'm running openclaw via coolify which has the model set via env var

What is the best way to upgrade? Right now I'm thinking:

Upgrade first to 5.10 then change env var from openai-codex/got-5.5, (not restart container yet so changes not applied) and then go to terminal and do doctor --fix and only after that's done restart

What do you think?

dreamy oreBOT May 12, 2026, 7:45 PM

#

I’d change the order a bit. Key point: doctor can repair OpenClaw config/state, but it cannot rewrite Coolify env vars. If the container process still sees openai-codex/gpt-5.5, that env override can keep winning.

Best path:

Back up your persistent OpenClaw volume/config first.
In Coolify, upgrade the image/package to 5.10-beta.3+ / 5.10 stable when available.
In the same Coolify env edit, change the model env var from:

openai-codex/gpt-5.5

to:

openai/gpt-5.5

#

Also make sure the Codex harness is enabled / runtime is set to codex somewhere in config/env. The important target is:

plugins: { entries: { codex: { enabled: true } } },
agents: { defaults: { agentRuntime: { id: "codex" } } }

Redeploy/restart the container so the new version + new env are actually active.
Exec into the new container and run:

openclaw doctor --fix
openclaw config validate
openclaw models status

Start a fresh chat/session and check /status — existing sessions can keep old PI/runtime pins.

#

I would not do “change env but don’t restart, then run doctor” unless you’re sure the shell running doctor already has the new env. In Coolify, usually it won’t; the old container/process still has the old env, and doctor may be looking at stale effective config.

Also tiny typo check: you wrote openai-codex/got-5.5 — make sure it’s gpt, not got 🦞

stable wave May 13, 2026, 1:10 AM

#

Okay so this is the order?

Backup

change env var

Update + restart

Change env var

Restart

Doctor --fix

Test and see if it works

Optional:

Come back, update to codex harness

See if it works

#

Okay so this is the order?

Backup

change env var

Update + restart

Change env var

Restart

Doctor --fix

Test and see if it works

Optional:

Come back, update to codex harness

See if it works

dreamy oreBOT May 13, 2026, 1:10 AM

#

Close — I’d simplify it to avoid two env changes/restarts.

Recommended order:

Backup your persistent OpenClaw volume/config.
In Coolify, change the env var once:

openai-codex/gpt-5.5 → openai/gpt-5.5

In the same Coolify deploy/update, upgrade to 5.10-beta.3+ / 5.10 stable and restart/redeploy.
Exec into the updated container and run:

openclaw doctor --fix

#

openclaw config validate
openclaw models status

Start a fresh session/chat and test.

For the Codex harness part: if your goal is to switch to harness, don’t leave it as a later mystery step — enable it during the same config pass if you can:

plugins: { entries: { codex: { enabled: true } } },
agents: { defaults: { agentRuntime: { id: "codex" } } }

Then after restart, /status should show Runtime: OpenAI Codex.

If you want the safest staged migration: first upgrade + switch model ref + doctor, confirm PI still works. Then enable agentRuntime: codex and restart/test. That’s slower but easier to debug if something breaks.

#

Almost — remove the duplicate env change/restart.

Do either the safe staged path:

Backup
Update to 5.10
Change env var once: openai-codex/gpt-5.5 → openai/gpt-5.5
Redeploy/restart
Run:

openclaw doctor --fix
openclaw config validate
openclaw models status

Test a fresh session

#

Then optionally enable Codex harness + restart + test again

Or the one-shot path:

Backup
Update to 5.10
Change env var to openai/gpt-5.5
Enable Codex harness/runtime in config
Redeploy/restart
doctor --fix + validate + status
Test fresh session

I’d recommend the staged path if this is production-ish. First confirm the model-ref migration works, then flip the harness switch.

stable wave May 13, 2026, 1:14 AM

#

Cool thanks will try and get back

dreamy oreBOT May 13, 2026, 1:14 AM

#

Sounds good — good luck with the migration 🦞

If it gets weird, paste openclaw --version, openclaw models status, and the relevant redacted env/config bits, and we can untangle it.

stable wave May 14, 2026, 2:25 PM

#

so only in 5.10 this naming thing changes right?

#is openai-codex?* not a thing anymore? as a model name?