LOCK

OpenAI Codex CLI with Meridian

Route the Codex CLI agent through Meridian so every prompt is authenticated, logged, and governed by your workspace policies. No direct OpenAI keys on developer machines.

Default model

gpt-5-codex

All requests flow through https://meridian.getnimbus.net/api/v1. The model parameter is injected server-side; clients cannot override it.

  1. 1

    Install Codex CLI

    npm install -g @openai/codex
  2. 2

    Set your Meridian API key

    export OPENAI_API_KEY=nim_live_***
  3. 3

    Point Codex at Meridian

    export OPENAI_BASE_URL=https://meridian.getnimbus.net/api/v1
  4. 4

    Run a refactor

    codex 'refactor this'

What happens next

  • · Codex sends its chat-completion requests to Meridian instead of api.openai.com.
  • · Meridian validates the bearer token, applies your workspace rate limits, and forwards the request to the upstream model.
  • · Responses are streamed back through the same connection — latency overhead is typically under 15 ms.

Need a provisioned API key? Generate one in the dashboard.