Tokens
How Solitud meters AI usage and how your wallet works.
What is a token?
A token is the unit Solitud uses to meter the AI work that builds your app. Whenever the AI reads your project, plans a change, writes code, and verifies the result, it consumes tokens — roughly in proportion to how much thinking and writing the request required.
Tokens cover the intelligence, not the hosting. Running your app on managed infrastructure is billed separately through your workspace resource tier.
How usage is metered
Each turn — one request and the work it triggers — debits tokens from your balance. A few things shape the cost of a turn:
- How much the AI reads. Larger apps mean more context to load before acting.
- How much it writes. Bigger or more complex changes produce more output.
- The model you choose. More capable models cost more per token (see below).
Token-weighted billing. Not all tokens cost the same to produce. Output and freshly-reasoned tokens are weighted more heavily than cached reads, so your balance tracks the true cost of the work rather than a flat count.
Your wallet
Your account holds a Token Wallet, shared across every workspace you own. Building draws it down; you top up to add more. You can watch the balance move per turn, so there are no surprises — a turn shows what it cost when it completes.
Top-ups start at $100, with bonus tokens on larger amounts:
| Top-up | Bonus |
|---|---|
| $100 | — |
| $500 | +10% |
| $1,500 | +25% |
Each Solitud AI keeps its own balance, so the exact token amount a top-up adds depends on the model and is shown in the app at top-up time. A starting balance (about one $100 top-up) is also what unlocks creating a workspace.
Choosing a Solitud AI
A Solitud AI is the model that does the building. Solitud offers a curated selection across a speed/cost vs. capability range, so you can match the model to the job:
| Tier | Good for | Relative cost |
|---|---|---|
| Budget | Small, well-defined changes; quick edits | Lowest |
| Standard | Everyday building and iteration | Moderate |
| Premium | Large features, tricky logic, big refactors | Highest |
Exact model names, availability, and per-token rates are shown in the app when you pick a Solitud AI for a workspace, and may change as better models ship.
Pick a cheaper model for routine tweaks and a more capable one when a change is ambitious — you'll get more reliable results for the harder work.
Estimating cost
A rough mental model:
- Tiny edits (copy change, a single field) — very cheap.
- A new feature (a screen plus its backend) — moderate.
- A big build from scratch or a large refactor — the most tokens, but delivered as a usable, running result.
The most efficient way to spend tokens is small, specific requests on a model that fits the task. Sprawling, vague prompts cost more and build less reliably.
FAQ
Do unused tokens expire? Your balance is yours to spend — check the dashboard for any current terms.
Does hosting use tokens? No. Tokens pay for AI work. Keeping your app running is part of your workspace resource tier.
What if a build fails partway? You're billed for the work the AI actually did. Building incrementally keeps any single turn small and cheap.