Troubleshooting

Quick lookup

The app says	Almost certainly	Fix
`401 Unauthorized`, `Invalid token`	Wrong key, missing characters, expired, or disabled	Paste the key into the check page
`404 Not Found`	Base URL missing `/v1` (or doubled to `/v1/v1`)	Fix the Base URL
`No available channel for model ... under group ...`	A key of one package calling the other package’s model	Use the model of your package, see Models
`model not found`, instant model-name error	Typo in the model name	Copy the name from the model table
`429 Too Many Requests`	Over the key’s per-minute request limit	Wait for the next minute, slow down
`insufficient quota`	Key has used all its tokens	Buy a new code
Config edited but nothing changes	App not fully quit, or terminal not reopened	Quit fully and reopen

Keys and quota

Wrong or truncated key

The most common problem is not a broken key but a few characters lost while selecting text with the mouse. Always use the Copy button on the redeem page. Paste the key into the check page: if it says valid, the key is fine and the problem is in the app configuration.

Key expired

Every package has a validity period counted from redemption (printed on the product). After expiry every request returns 401 even if quota remains. The check page shows the expiry date. Unused tokens do not transfer to a new code.

Quota used up

Everything worked, then suddenly every request fails with insufficient quota. See what is left on the check page. Quota dropping unusually fast is usually an agent reading the whole repo, or a very long conversation being resent every turn.

Lost key

Redeem the same CDK code again at redeem.teamsoclo.site; the page shows the original key. Codes are not re-issued with a new key.

How many machines per key

There is no device limit, but the per-minute request limit is shared by every machine using the key. Sharing a key means users compete for the same limit and hit 429 sooner.

Connection

Missing /v1

The correct Base URL is https://gpt.teamsoclo.site/v1. Without /v1 requests hit the website instead of the API and the app reports 404 or “could not connect”.

The opposite also fails: some apps (Chatbox, some Cline builds) append /v1 themselves. If you entered it and still get 404, try without /v1; the real address may have become /v1/v1.

Calling the other package’s model

No available channel for model gpt-astra under group default means a GPT key is calling the Astra model (or the reverse). Each key can only call its own package’s model. Verify with:

curl https://gpt.teamsoclo.site/v1/models -H "Authorization: Bearer sk-xxx"

429 Too Many Requests

Each key has a maximum number of successful requests per minute set by its package. Above that you get 429 with a “request limit reached” message. Wait for the next minute. For batch scripts add a delay between calls, or buy another code to spread the load.

Reply cut off mid-way

For questions that need long thinking, a non-streaming request can be dropped before the first token. Enable stream: true. Most chat apps already do; add it when calling from code.

`Unsupported parameter: reasoning_effort`

Only Chat Completions accepts a top-level reasoning_effort. With the Responses API use "reasoning": {"effort": "high"}. Codex handles this itself.

Codex

`codex: command not found` after installing

npm finished but the current terminal has not reloaded PATH. Open a new terminal. If it still fails, check that npm’s bin folder is on PATH:

npm config get prefix

That folder plus /bin must be in $PATH. The TEAMSOCLO_API_KEY variable has not reached the Codex process:

Terminal: open a new tab, check echo $TEAMSOCLO_API_KEY.
VS Code on macOS: quit with Cmd + Q and reopen. The install command set the variable for GUI apps via launchctl.
VS Code on Linux: start VS Code from a terminal (code .), or log out and back in.
Windows: open a new PowerShell or VS Code after setx.

Check that config.toml has requires_openai_auth = false and env_key = "TEAMSOCLO_API_KEY".

Edited config.toml but Codex does not change

Config is read only at process start. Reload Window is not enough; quit VS Code fully (macOS Cmd + Q, Windows close every window and background process). This is the most frequently asked Codex question.

`EACCES: permission denied` on npm install

Node was installed with system privileges, common on Linux. Change npm’s prefix:

npm config set prefix ~/.npm-global
export PATH="$HOME/.npm-global/bin:$PATH"

Add the export line to ~/.bashrc or ~/.zshrc. Long term, install Node via nvm, see Getting started.

Windows: `running scripts is disabled on this system`

PowerShell blocks .ps1 files, and codex / npm typed in PowerShell go through npm’s .ps1 shims. Run once:

Set-ExecutionPolicy -Scope CurrentUser RemoteSigned

Then open a new PowerShell. Or type codex.cmd, npm.cmd instead.

Windows: `EBUSY: resource busy or locked`

Codex is running (CLI, extension or app) so npm cannot overwrite codex.exe. Close everything running Codex and rerun the install command. The config was already written before this step; only the CLI upgrade was skipped.

Windows: config file encoding error

Editing with PowerShell 5.1 and Set-Content -Encoding UTF8 adds a BOM that Codex cannot read. Use the install command again (writes UTF‑8 without BOM), or save the file in VS Code as UTF-8 (not UTF-8 with BOM).

Restore the previous config

Each run of the install command backs up the old config next to the new one:

ls ~/.codex/*.bak.*
cp ~/.codex/config.toml.bak.20260925143000 ~/.codex/config.toml

With the tool, choose Reset and it restores automatically.

Frequently asked

How do I change keys

Rerun the Codex install command with the new key, or rerun the tool. For Cursor, Cline and Chatbox, edit the API Key field in Settings.

Is it safe to run the install command several times

Yes. It is designed for reruns: the old config is saved as .bak.<timestamp>, the CLI is upgraded, and the old key line in your shell rc is replaced.

Does the command send my key anywhere

No. You paste the key into the command on your own machine; it only writes an environment variable and a local config file. The entire command is shown on the page, read every line before running.

Remove everything

See Uninstall on the Codex page. For other apps just delete the key in the app’s Settings.

Still not working

Is the key alive

Paste it into the check page. An error here means the key (wrong, expired, out of quota), not the app.

Call the API directly with cURL

curl https://gpt.teamsoclo.site/v1/chat/completions \
  -H "Authorization: Bearer sk-xxx" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-5.6-sol","messages":[{"role":"user","content":"hi"}]}'

If this works and the app does not, the problem is the app configuration.

Recheck three things in the app

Base URL ends with /v1, the key is complete, the model matches your package. These cause nearly every issue.

Contact support

Send a screenshot of the error and the last 4 characters of your key to t.me/teamsoclo. Never send the full key.

Getting started

Setup

Support

Quick lookup

Keys and quota

Wrong or truncated key

Key expired

Quota used up

Lost key

How many machines per key

Connection

Missing /v1

Calling the other package’s model

429 Too Many Requests

Reply cut off mid-way

`Unsupported parameter: reasoning_effort`

Codex

`codex: command not found` after installing

Edited config.toml but Codex does not change

`EACCES: permission denied` on npm install

Windows: `running scripts is disabled on this system`

Windows: `EBUSY: resource busy or locked`

Windows: config file encoding error

Restore the previous config

Frequently asked

How do I change keys

Is it safe to run the install command several times

Does the command send my key anywhere

Remove everything

Still not working

​Quick lookup

​Keys and quota

​Wrong or truncated key

​Key expired

​Quota used up

​Lost key

​How many machines per key

​Connection

​Missing /v1

​Calling the other package’s model

​429 Too Many Requests

​Reply cut off mid-way

​Unsupported parameter: reasoning_effort

​Codex

​codex: command not found after installing

​Codex still asks for an OpenAI login

​Edited config.toml but Codex does not change

​EACCES: permission denied on npm install

​Windows: running scripts is disabled on this system

​Windows: EBUSY: resource busy or locked

​Windows: config file encoding error

​Restore the previous config

​Frequently asked

​How do I change keys

​Is it safe to run the install command several times

​Does the command send my key anywhere

​Remove everything

​Still not working

Quick lookup

Keys and quota

Wrong or truncated key

Key expired

Quota used up

Lost key

How many machines per key

Connection

Missing /v1

Calling the other package’s model

429 Too Many Requests

Reply cut off mid-way

`Unsupported parameter: reasoning_effort`

Codex

`codex: command not found` after installing

Codex still asks for an OpenAI login

Edited config.toml but Codex does not change

`EACCES: permission denied` on npm install

Windows: `running scripts is disabled on this system`

Windows: `EBUSY: resource busy or locked`

Windows: config file encoding error

Restore the previous config

Frequently asked

How do I change keys

Is it safe to run the install command several times

Does the command send my key anywhere

Remove everything

Still not working