Codowave
Sign inGet started
FAQ
FAQ

Which LLMs Does Codowave Support?

Codowave runs on your own LLM subscription. Claude Code, Codex, and the major hosted models work out of the box, with per-task model routing so the planner, coder, and reviewer can use different brains.

2 min read

Which LLMs Does Codowave Support?

Codowave is bring-your-own-key. Connect a subscription, the agent uses it. Different roles (planner, coder, reviewer) can route to different models, so you can pair a cheap reasoning model for planning with a strong coding model for diffs.

Start your 5-day trial

Supported today

  • Claude Code — recommended for the coder and reviewer roles. Best coding diffs in our benchmarks.
  • Codex (GPT-4.1+) — strong for the planner role. Connects via your OpenAI key.
  • Anthropic API direct — for teams already paying for Claude through the API.
  • OpenAI API direct — GPT-4o, GPT-4.1, and o-series reasoning models.
  • Local + self-hosted — anything with an OpenAI-compatible endpoint (Ollama, LM Studio, vLLM). Useful for offline / regulated environments.

Per-role model routing

The pricing config lets you assign a different model per agent role:

  • Planner — usually a cheap reasoning model (o3-mini, Sonnet)
  • Coder — the strongest coding model you have access to (Claude Code, GPT-4.1)
  • Reviewer — same family as the coder, with reasoning enabled

You can also pin a single model across all roles if you don't want the complexity.

Frequently asked questions