Choose your AI engine
Claude runs your sessions by default. Turn on other engines, such as Codex, Gemini or DeepSeek, and pick one per session or per project.
Every session runs on an AI engine. Claude (Claude Code) is the default, and you can also run sessions on other engines, such as OpenAI's Codex, Google's Gemini, or cheaper pay-per-use models like DeepSeek. They share the same sessions list, Inbox and chat.
Turn on other engines#
Open Settings > Accounts.
Turn on Show alternative AI providers.
Under Other AI tools, open the engine you want and follow its card: install its program, sign in or paste a key, and turn on its Allow ... sessions in any project switch.
Each card says what is missing. The engine pages below walk through the common ones.
Pick the engine for one session#
Start a new session.
Before your first message, click the engine name next to the session title (it says Claude Code at first) and pick another one.
Send your message. The session keeps that engine from then on.
An engine that still needs setting up shows Set up, Install or Add key in that list; click it to open the card that fixes it.
DeepSeek, Kimi, GLM, MiniMax and Meta run inside Claude Code: keep Claude Code and pick one of their models in the model choice next to it.
Set a default engine#
For one project: right-click it, choose Edit Project > More options, and pick a Default provider.
For every project: pick the Default Provider in Settings > Sessions.
Who pays, and what works#
| Engine | You pay with | Asks before risky steps |
|---|---|---|
| Claude Code | Your Claude plan or Anthropic key | Yes, by your permission level |
| Codex | Your ChatGPT plan, or an OpenAI key | Yes, by Codex permission level |
| Gemini | A Gemini API key (see its page) | No |
| Anti-Gravity | Your Google AI plan (Google sign-in) | No |
| DeepSeek, Kimi, GLM, MiniMax, Meta | Your key with that company, per use | Yes, like Claude |
| OpenCode | Your Anthropic or OpenRouter key, per use | No |
| Cursor, Grok Build, Hermes, Pi | Your account or key with that tool | No |
"No" means the engine runs its tools (commands, file edits) on its own once you start it.
Spending stops that are already on#
Sessions that bill a key for every message have spending stops from the start. You find them in Settings > Sessions > Spending limits:
Budget Caps (per session): a warning when one session has spent $20, and the session stops at $100. This covers sessions on an Anthropic API key and on DeepSeek, Kimi, GLM, MiniMax and Meta keys.
Daily limit (per account): a warning when an Anthropic API key has spent $10 today, and at $25 its running sessions stop and new ones wait until tomorrow.
Daily limit (per metered provider): a daily stop for each of DeepSeek, Kimi, GLM, MiniMax and Meta. It is off until you set it.
Sessions on your Claude plan are never stopped by these limits, and neither are engines billed elsewhere, such as Codex and Gemini. Change any amount there, or clear a box to turn that stop off.
Cost figures. Statistics shows spending per engine. Claude, OpenCode and Pi report exact costs; Codex, Cursor and the pay-per-use engines are estimates; Gemini, Anti-Gravity, Grok Build and Hermes report none.
Dedicated sidebar items. Codex, Gemini and Anti-Gravity can each add their own item to the projects column (a Show ... virtual project switch on their card). Sessions started there always use that engine.
Related
- Keep long sessions affordableEvery reply re-reads the whole conversation. Yardmaster warns you when a session gets large, holds costly automatic replies, and can continue in a fresh session.
- Run sessions on Gemini or Anti-GravityUse Google's models in Yardmaster: Gemini with an API key, or Anti-Gravity with your Google AI plan.
- Run sessions on DeepSeek, Kimi, GLM, MiniMax or MetaThese engines run inside Claude Code with your own key from each company, often for less money. Turn one on and pick its model.