Integrations

OpenAI Codex CLI

The OpenAI Codex CLI is a coding agent that runs in your terminal.

Updated
Reading time
3 min

On this page

You can point it at Apertis and run it on open-weight models as well as Claude, Gemini and GPT models.

Prerequisites

  • Node.js 18 or later, for the npm install
  • An Apertis API key, from the Apertis Dashboard

Step 1: Install Codex

npm install -g @openai/codex
codex --version

See the Codex documentation for other install methods.

Step 2: Add Apertis as a provider

Codex reads custom providers from ~/.codex/config.toml. Copy this block into that file:

model = "glm-5.1"
model_provider = "apertis"

[model_providers.apertis]
name = "Apertis"
base_url = "https://api.apertis.ai/v1"
env_key = "APERTIS_API_KEY"
wire_api = "responses"
  • base_url is https://api.apertis.ai/v1. It must end in /v1.
  • env_key names the environment variable that holds your key. Codex reads the key from it, so the key never goes into the file.
  • wire_api = "responses" makes Codex send its requests in the OpenAI Responses format, which Apertis accepts at https://api.apertis.ai/v1/responses. When a model's provider does not serve the Responses endpoint, Apertis converts the request and the response. See the Responses API reference.

Step 3: Set your API key

export APERTIS_API_KEY="your_apertis_api_key"

Add the line to your shell profile (~/.zshrc or ~/.bashrc) to keep it across sessions.

Step 4: Choose a model

Set model in config.toml, or pass -m <model> on the command line. These models completed a Codex editing task through Apertis:

Model ID Developer
glm-5.1 Z.AI (GLM), open weights
kimi-k2.6 Moonshot AI (Kimi), open weights
deepseek-v3.2 DeepSeek, open weights
minimax-m2.5 MiniMax, open weights
claude-sonnet-4-6 Anthropic
gemini-3.1-pro-preview Google
gpt-5.5 OpenAI

Each of these models completed the same task three times out of three with Codex CLI 0.166.0 on 2026-11-10: fix a bug in a file, then write a second file.

For every model you can call, see the Model List.

Verify the connection

# 1. The key is valid: lists the models it can call
curl -s https://api.apertis.ai/v1/models \
  -H "Authorization: Bearer $APERTIS_API_KEY" | head -c 300

# 2. Codex answers through Apertis
codex exec --skip-git-repo-check "Reply with the word ok"

The second command ends with the model's reply, ok. Then run codex in a project directory to start an interactive session.

Troubleshooting

Symptom Fix
401 or Invalid API Key Check that APERTIS_API_KEY is exported in the shell that runs codex, and copy the key again from the Dashboard.
model not found or 404 Use a model ID exactly as it appears in the Model List, and make sure base_url ends in /v1.
Codex still uses its default provider Check that model_provider = "apertis" is at the top level of config.toml, above the [model_providers.apertis] table.