Skip to main content
Codex CLI supports custom model providers through ~/.codex/config.toml, so you can route requests to AI Stats without changing your local workflow. If you also want Codex to read the live AI Stats documentation while it works, add the Docs MCP server as well.
Need reusable agent setup help as well as provider wiring? Install the public AI Stats skills from AI-Stats/skills:
npx skills add AI-Stats/skills --agent codex --skill '*'

​
Prerequisites

  • An AI Stats API key.
  • Codex CLI installed locally.

​
1) Set your API key

export AI_STATS_API_KEY="your_ai_stats_api_key"

$env:AI_STATS_API_KEY = "your_ai_stats_api_key"

​
2) Configure Codex to use AI Stats

Create or update ~/.codex/config.toml: 
model = "openai/gpt-5-3-codex-2026-02-05"
model_provider = "ai_stats"

[model_providers.ai_stats]
name = "AI Stats"
base_url = "https://api.phaseo.app/v1"
env_key = "AI_STATS_API_KEY"
wire_api = "responses"

​
3) Start Codex

codex

​
Verification

  • Confirm your model id exists in Models before launching.
  • Use Responses for the default Codex-style flow.
  • If needed for compatibility, switch wire_api to chat and use Chat Completions.

​
Troubleshooting

  • 401 Unauthorized: check that AI_STATS_API_KEY is set in the same shell running codex.
  • model_not_found: fetch a current id from Models and update model.
  • Provider-specific issues: follow Routing and fallbacks for failover behavior.
Last modified on May 19, 2026