“We need multi-model support.”
— Just about everyone
We heard you. Good news: Cate was never just about Claude.
It's been multi-model since its early days. Early access builds are Claude-only, but internal builds run Claude, Codex, and OpenCode, covering dozens of models. We've been quietly building toward a world where you pick the right model for the right job, not the right model for every job.
My ideal workflow? Opus for planning, with a Codex double-check. Grok for writing code. Three different models, three different strengths, one coordinated workflow.
Your Rig, Your Budget
The vision: you pick the models, you set the budget envelope, Cate routes the work.
Planning agent runs Opus. Implementation agents fan out across Grok, Sonnet, whatever fits. Review runs Codex. Each model bills on its own meter. You're not choosing between a good plan and a cheap implementation. You're getting both.
The bigger idea: bring the tool you want into the workflow you already know. That's why Cate exists. Your SDLC doesn't change. Your issue tracker doesn't change. You just get to pick which engine does the work. I've been wanting to try running the OpenCode runtime with OpenRouter's auto-mode as a provider, letting it pick the best model per task automatically, but I haven't had the time. That's the point, though. The experiment is possible because the orchestration layer doesn't care which model is behind the curtain. Same flow, same board, same PR process, different engine.
One note on Claude billing: Anthropic is separating Agent SDK usage from interactive Claude Code starting June 15, with a monthly “credit” for SDK use on Pro and Max plans. Cate will support both paths: Agent SDK and raw interactive Claude CLI sessions, so you choose what makes sense for your plan and your workflow. If you're on a Pro or Max subscription and want to get your full money's worth through interactive sessions, Cate won't force you onto the SDK. Your rig, your rules.
Hopefully Next Week
Multi-model support is targeting early access next week. Codex, OpenCode, and dozens of models behind them, alongside the Claude you're already using.
If you want to be the person running Opus for planning and Codex for review while cheap models handle volume, request early access or get in touch. I want to hear how you'd split the work.
Ready to put your Claude Code setup to work? Request early access to Cate, or get in touch. We'd love to hear what you're building.