Credits: the fuel for AI features
Every AI action on Spinini — agent runs, chat messages, code suggestions — consumes credits. Credits are shared across the AI Agent (Spinini AI) and the AI Chat panel. Everything else — the editor, terminal, file management, running your container, publishing — is free.
Your credit balance is shown in the top-right corner of your dashboard and in the editor header at all times.
When credits are consumed
AI Agent runs — the biggest consumer. Each run costs credits based on how many tokens the model processes: the files it reads, the code it writes, and the conversation history. A simple one-file edit might cost 20–50 credits. Building a full feature from scratch might cost 200–500.
AI Chat messages — lighter than agent runs. A question-and-answer exchange typically costs 10–40 credits. The conversation history grows over time, so very long threads cost more per message. Start a new conversation if an old thread is getting very long.
Model choice matters — DeepSeek R1 uses more credits than V3 because of its reasoning chain. Spinini Deep costs more than Spinini Smart. Same task, different cost depending on model.
What doesn't consume credits
- Editing files yourself
- Running your container
- Terminal commands
- Publishing your app
- Live preview
- Loading the dashboard or any page
- Browsing the Gallery
Top-up packages
You can buy more credits at any time — top-ups don't expire and stack on top of your plan's monthly allocation.
| Package | Credits | Best for |
| $5 | 500 credits | A quick top-up to finish a session |
| $9 | 1,000 credits | A few days of moderate AI use |
| $22 | 3,000 credits | A week of heavy building |
| $65 | 10,000 credits | A month of intensive development |
Purchase from the Billing page or directly from the "Out of Credits" modal that appears when you hit zero.
The out-of-credits experience
When you run out, a modal appears explaining the situation and showing the top-up packages. You can purchase right there without leaving the editor — no need to go find the billing page. After purchasing, credits are added instantly and you can continue your agent run immediately.
Tips to make credits go further
Use DeepSeek V3 for simple tasks. V3 is significantly cheaper than R1 and handles quick questions, small edits, and simple fixes perfectly well. Switch to R1 only when you're working on something genuinely complex.
Use AI Chat for questions. If you want to understand code or ask a conceptual question, use the Chat panel rather than the Agent. Chat doesn't read your project files unless you paste them in, which makes it cheaper.
Be specific in your prompts. Vague prompts ("make this better") cause the agent to read many files, think broadly, and write a lot. Specific prompts ("add a loading spinner to the submit button in src/components/LoginForm.tsx") focus the agent and cost a fraction as much.
Keep file context focused. When you ask the agent to work on something, reference specific files. "Update the function on line 45 of api/auth.js" is cheaper than "update the auth logic" because the agent reads exactly one file instead of scanning the whole project.
Watch your balance. The editor shows a low-credit warning when you drop below 20 credits. Heed it — it's easy to lose track in the middle of a long session.
Use checkpoints before big runs. Not a credit tip per se, but it saves you from having to re-run a failed agent task twice.