Codex CLI + Nebius Token Factory

Setup and configure Codex CLI for AI-assisted coding using Nebius Token Factory models over an OpenAI-compatible API.

Quick Start

# 1. Install Codex CLI
npm install -g @openai/codex

# 2. Download and run the setup script
curl -fsSL https://raw.githubusercontent.com/opencolin/codex-nebius/main/setup-codex-nebius.sh -o setup-codex-nebius.sh
chmod +x setup-codex-nebius.sh
bash setup-codex-nebius.sh

# 3. Start coding
codex "Write a Python function to validate email addresses"

Tip: Download the script first instead of piping directly to avoid corruption issues.

Features

🔄 Flexible — Switch models instantly via configuration
🚀 Fast Setup — Automated configuration script
📦 Three Profiles — Fast, balanced, and precise models
🎯 Unified CLI — One interface for multiple LLM providers

What's Included

• setup-codex-nebius.sh — Automated setup script

  • Checks dependencies
  • Creates ~/.codex configuration
  • Prompts for API key securely
  • Tests connection to Token Factory
  • Generates pre-configured profiles

• CODEX_CLI_SETUP_GUIDE.md — Complete setup guide

  • Manual configuration steps
  • Advanced profile customization
  • Integration with editors (VSCode, Neovim)
  • Git workflow integration
  • Troubleshooting & security best practices

Prerequisites

• Codex CLI — Install with npm install -g @openai/codex
• Nebius Token Factory API key — Get from https://nebius.com/
• bash or zsh — For the setup script

Usage

Basic Commands

# Using default profile (balanced)
codex "Write a Python function"

# Using fast profile
codex --profile nebius-fast "Quick implementation"

# Using precise profile
codex --profile nebius-precise "Detailed explanation"

Override at Runtime

# Change temperature for more/less creative output
codex -c temperature=0.3 "Be precise"

# Limit response length
codex -c max_tokens=1024 "Short answer"

Profiles

Profile	Model	Use Case	Speed
nebius-token-factory (default)	Hermes-4-405B	Balanced, all-purpose	Medium
nebius-fast	Gemma-3-27b	Quick suggestions and completions	Fast
nebius-precise	Qwen3-Coder-480B	Complex coding and analysis	Slow

Configuration

After running the setup script, configuration is stored in ~/.codex/config.toml. All three profiles share a single nebius_token_factory provider:

default_provider = "nebius_token_factory"
default_profile = "nebius-token-factory"

[model_providers.nebius_token_factory]
name = "Nebius Token Factory"
base_url = "https://api.tokenfactory.nebius.com/v1"
env_key = "NEBIUS_API_KEY"
wire_api = "responses"

# Balanced (default)
[profiles.nebius-token-factory]
model_provider = "nebius_token_factory"
model_name = "nebius/NousResearch/Hermes-4-405B"
max_tokens = 4096
temperature = 0.7

# Fast
[profiles.nebius-fast]
model_provider = "nebius_token_factory"
model_name = "nebius/google/Gemma-3-27b-it"
max_tokens = 2048
temperature = 0.5

# Precise
[profiles.nebius-precise]
model_provider = "nebius_token_factory"
model_name = "nebius/Qwen/Qwen3-Coder-480B-A35B-Instruct"
max_tokens = 8192
temperature = 0.3

See CODEX_CLI_SETUP_GUIDE.md for detailed configuration options.

Upgrading from 1.x

v2.0 is a breaking release: the redundant nebius_fast provider was removed and all three profiles now reference the single nebius_token_factory provider. If you have an existing ~/.codex/config.toml from a 1.x setup, just re-run the setup script — it backs up your old config to config.toml.bak.<timestamp> before writing the new layout. See the Migration notes in the CHANGELOG for details.

Troubleshooting

"bash: line XXX: cho: command not found"

This error occurs when the setup script is corrupted during download via pipe.

Solution: Download the script first, then run it:

curl -fsSL https://raw.githubusercontent.com/opencolin/codex-nebius/main/setup-codex-nebius.sh -o setup-codex-nebius.sh
chmod +x setup-codex-nebius.sh
bash setup-codex-nebius.sh

Or clone the repository:

git clone https://github.com/opencolin/codex-nebius.git
cd codex-nebius
bash setup-codex-nebius.sh

"Codex CLI not found"

npm install -g @openai/codex

"UNAUTHENTICATED" error

Verify your Nebius API key:

echo $NEBIUS_API_KEY

Reload your shell profile if you just set it:

source ~/.zshrc  # or ~/.bashrc

Connection test failed

Check connectivity to Token Factory:

curl -H "Authorization: Bearer $NEBIUS_API_KEY" \
  https://api.tokenfactory.nebius.com/v1/models

For more troubleshooting, see CODEX_CLI_SETUP_GUIDE.md.

Resources

• Codex CLI GitHub — Official Codex CLI repository
• Nebius Documentation — Official Nebius docs
• Nebius Token Factory — Token Factory service
• Codex Provider Configuration — Provider setup guide

License

MIT

Contributing

Found an issue or have suggestions? Please open an issue or submit a PR!

---

Part of the codex-nebius project — AI Skills for Nebius infrastructure