Auto mode
The right model for every step of the task.
Pick Auto mode once and stop choosing models. Capuchin does the work, a frontier model steps in when it changes the outcome, and you never hit a limit wall in the middle of a task.
Task: add rate limiting to file uploads
- 1Capuchin
Read the upload route and 4 related files
src/routes/upload.ts - 2Capuchin
Added rate-limit middleware and wired it in
src/middleware/rateLimit.ts - 3Capuchin
Ran the tests: 3 failed
- 4Capuchin
Tried a fix. Same 3 tests still fail
- 5Frontier · tests failed
Found the cause: a shared clock mock leaking between tests
- 6Capuchin
Applied the fix. 42 of 42 tests pass
When the frontier model steps in
Auto watches real signals from the run, not a guess about the prompt. When one fires, the frontier model gets a short handoff summary of the task so far, not the whole conversation again, so the switch is quick and cheap. Once the problem is solved, Capuchin takes over again.
- Tests keep failing
- An edit adds new errors
- The agent goes in circles or stops making progress
- You reject or revert its changes
- You correct it
- The task looks hard from the start
Two profiles
Auto mode (Balanced)
Capuchin does the work. The frontier model steps in only when Capuchin is struggling, then hands back. The best value for everyday work.
Pro, Team and up
Auto mode (Max quality)
The frontier model plans every task and reviews the result. Capuchin does the work in between. For changes where a second pair of eyes is worth it.
Pro+, Ultra, Enterprise
Every step is tagged
Each reply shows the model that ran it: Capuchin, or the frontier model with the reason, like tests failed or reviewing. In the IDE, hover the tag to see how often the frontier model stepped in on this task.
You can overrule it
Already know a task is hard? Press the rocket button in the IDE or Code Agent, or type /frontier in the CLI. The frontier model stays on for that conversation.
No limit walls
Frontier steps draw on your Premium pool. If it runs out mid-task, Auto carries on with Capuchin, tells you, and brings the frontier model back when the pool resets.
In every app, on one engine
- IDE
- Model picker in the chat panel → Auto mode (top of the list)
- Code Agent
- Model menu in the prompt bar → Auto mode
- CLI
/model auto-modein a session, ormc --model auto-mode- Web IDE and cloud jobs
- Model menu → Auto mode
Needs IDE 1.3.0, Code Agent 1.1.0 or CLI 1.1.0. Download or update.
Plans
| Plan | Balanced | Max quality |
|---|---|---|
| Free | ||
| Pro, Team | ||
| Pro+, Team Pro+ | ||
| Ultra, Team Ultra, Enterprise |
The Capuchin promo trial includes Capuchin, not Auto. See pricing.
Questions about Auto mode
What is Auto mode in MonkeysCode?
Auto mode picks the model for each step of a coding task. Capuchin, the MonkeysCode coding model, runs the loop. A frontier model steps in only when the task needs it: tests keep failing, an edit adds errors, the agent goes in circles, or you correct it. Once the problem is solved, Capuchin takes over again.
Which frontier model does Auto mode use?
The frontier model can change as better ones come out. The model picker always shows the current one, and every step is tagged with the model that actually ran it.
What is the difference between Auto mode and Max quality?
Auto mode (Balanced) brings in the frontier model only when Capuchin is struggling. Max quality has the frontier model plan every task and review the result, with Capuchin doing the work in between.
What happens if I run out of frontier usage mid-task?
Auto doesn't stop. It carries on with Capuchin, tells you so, and uses the frontier model again once your Premium pool resets. Dashboard → Usage shows how much is left and when it resets.
Can I force the frontier model?
Yes. Use the rocket button in the IDE or Code Agent, or /frontier in the CLI. It applies to that conversation only and ends when you switch model or clear the conversation.
Which plans include Auto mode?
Auto mode (Balanced) is on Pro and Team. Max quality is on Pro+, Ultra, their Team versions, and Enterprise. Free and the Capuchin promo trial don't include Auto. Your own provider key doesn't unlock Auto, because Auto routes between models that MonkeysCode runs.
Can I use Auto mode in scripts and CI?
Yes. Pass --model auto-mode (or auto-max) to mc -p. With --output-format stream-json, each Auto request also emits a route event with the model that ran and why.
Let Auto mode pick. Keep your focus on the code.
Auto mode comes with Pro, alongside unlimited Capuchin and a frontier allowance for Claude, Gemini and ChatGPT.