How to configure OpenAI Codex with Earbowe API

Updated 2026-07-24 · Technical setup guide.

Updated 2026-07-24 · Technical setup guide (not a sales post).

You will set two values in your coding tool:

• Base URL: https://api.earbowe.com/v1
• API key: from Earbowe console

China direct alternative: https://api.3861343.xyz/v1

1. Get a key

1. Sign in at api.earbowe.com/sign-in
2. Create a token
3. Confirm models with Models or GET /v1/models

2. Environment variables (works for many Codex / OpenAI-compatible CLIs)

export OPENAI_BASE_URL="https://api.earbowe.com/v1"
export OPENAI_API_KEY="sk-..."
# optional default model for tools that read it
export OPENAI_MODEL="gpt-5.5"

Then start your Codex CLI / agent in the same shell.

3. If the tool uses a config file

Many tools accept a provider block like:

{
  "model": "gpt-5.5",
  "baseURL": "https://api.earbowe.com/v1",
  "apiKey": "sk-..."
}

Field names vary (base_url, api_base, openaiBaseUrl). The important part is always:

• point to .../v1
• send Bearer token
• use a model ID returned by /v1/models

4. Verify before coding

curl -sS https://api.earbowe.com/v1/models \
  -H "Authorization: Bearer ***" | head

curl -sS https://api.earbowe.com/v1/chat/completions \
  -H "Authorization: Bearer ***" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-5.5","messages":[{"role":"user","content":"ping"}]}'

5. Troubleshooting

• 401 — key wrong, or not exported in the terminal that launched Codex.
• Model not found — switch to gpt-5.5 or another ID from /v1/models.
• Timeouts in China — try https://api.3861343.xyz/v1.
• Stream drops — disable extra HTTP proxies; use the API subdomain, not the www content site.

Open Console · Quickstart