Free
Begin with the essentials
$0/month
- 30 tasks a month
- Fast and balanced model presets
- Build and refine tasks with AI in the workbench
- Evals on every run. Know if it actually worked
- API access included
Prices that work with you
Grow
Do more with Sunnyday
$49/month
- 75 tasks a month
- Performance model preset unlocked
- Overage at $0.99 per extra task
- Set a monthly spend cap.
- Everything in Free
- Missing an integration? Paying users decide what I build next.
Pro
The take the day off package
$99/month
- 150 tasks a month
- Overage at $0.75 per extra task
- Choose your exact model
- Frontier models like Opus and Sol, at 5 tasks per run
- Custom evals
- Everything in Grow
Need more than Pro?
Custom volume, private deployments, integrations, and invoicing. Tell me what you're running and I'll build the setup around it.
Need help with your agents?
I consult on agent reliability: audits, evals, and hands-on help making agents work every time, not just in the demo.
Every plan includes
Build
- Unlimited agents — pay only when they run
- Plain-English authoring in the workbench
- Smart {{variables}} input detection
- Versioned agents
- Auto-generated names, descriptions & icons
- Your own files as context
Watch
- Live streaming runs
- Chat with a running agent
- Cancel instantly
- Reconnect without losing anything
- Full run transcripts
Models
- Most capable, balanced, and fast tiers
- Extended thinking for hard tasks
- Built-in web search
- Multimodal: agents see images and PDFs
Evaluate
- Evals on every run
- Pass / partial / fail scoring
- Output-quality grading
- Re-score any past run
- Per-run token & cost transparency
Run
- Fresh isolated sandbox per run
- A real computer, not a chat window
- Pandas, Excel, Word, PDF, Node included
- Reads and authors PDFs and Word docs
- Safety-checked file writes
- Plans before acting
- Self-correcting retries
Integrate
- Trigger by API, on a schedule, or with a click
- Streaming API responses
- File inputs & downloads over API
- GitHub: run on push, tag, or PR
- Team workspaces
Frequently Asked Questions
What counts as a task?
One task is one triggered run of an agent, from start to finish, no matter how many steps it takes. The only exception: runs on frontier models like Claude and GPT count as 5 tasks, because those models cost significantly more to operate.
What happens when I run out of tasks?
On the Free plan, your agents pause until your quota resets next month. On Grow and Pro, overage kicks in at your plan's per-task rate, up to a monthly spend cap you set yourself. You'll never get a bill you didn't see coming.
What models can I use?
Every plan runs on curated presets: Fast for quick jobs, Balanced for everyday work, and Performance (Grow and up) for the hard stuff. Presets are built from the best open models available and improve as better models ship, without you changing anything. On Pro you can also pin your exact model, including frontier models from Anthropic and OpenAI.
Do unused tasks roll over?
No. Your quota resets at the start of each billing cycle.
Is my data used to train models?
No. We don't train on your data, and your tasks are only routed to model providers that don't train on it either.
Can I change or cancel my plan anytime?
Yes. Upgrades apply immediately; downgrades and cancellations take effect at the end of your billing cycle. Your agents and run history stay put.