Eugene PlexusGitHub

Guides · v0.2.1

Use Claude Code or Codex with local models

Point Claude Code or Codex at your own local models, with one address and a key from the console.

  1. In the console’s Home page, find Use it from your apps.
  2. Under Key permissions and limits, fill in What is it for?, for example “Codex on the laptop”, and press Make a key. Copy it now: it is shown once and not stored.
  3. Choose the model, then press Codex or Claude Code under Set up.
  4. Put what it shows in place:
    • Codex: the block goes in ~/.codex/config.toml, and the key in the EUGENE_API_KEY environment variable, set where you start Codex. Never put the key in the file.
    • Claude Code: set the environment variables it lists: ANTHROPIC_BASE_URL, ANTHROPIC_AUTH_TOKEN, ANTHROPIC_MODEL, and, when the model’s window is known, its window and reply limit.
  5. Start Codex or Claude Code as usual.
Good to know
  • Choose a model that calls tools well. Protocol support does not make a small model a dependable coding agent.
  • Claude Code’s address has no /v1; Codex’s keeps it. The console writes both correctly.
  • A setting the model ignores is named in the x-eugene-plexus-ignored-settings response header.
  • A Claude Code turn reserves its whole answer allowance (32,000 tokens), so a llama.cpp server with a 64k-token pool runs one Claude Code turn at a time. The others wait; they do not fail.
  • Open WebUI, Continue, Cline and other apps have their own buttons under Set up.