How WormGPT works: tokens, models, studios and the API
A walkthrough of the four parts of the workspace and how usage is charged.
One balance, one account
Everything in WormGPT runs off a single account and a single token balance. Chat, the Studios, the developer API and the MCP server all draw down the same balance, so you never have to track separate quotas per feature. Access is $29.99 per month or $99.99 one-time for lifetime access. Full details are on the pricing page.
How charging works
Each request reserves an estimated amount of tokens before the model runs, then settles against the real usage once the response finishes. If a request fails part way through, the unused reservation is returned to your balance. That hold-and-settle cycle is why two concurrent requests can never push a balance below zero, and why the number you see on your dashboard matches what was actually consumed.
Every deduction, purchase credit and adjustment is written to a usage ledger you can review in your dashboard, so a balance can always be traced back to the events that produced it.
Model routing and fallback
Chat requests are routed across several supported models rather than pinned to one. If the first model is unavailable or times out, the request retries down a fallback chain instead of failing in your face. You can also pick an engine and a persona from the mode bar in chat when you want a specific behaviour, and attach screenshots, charts or documents for analysis in the same conversation.
Image and Video Studio
Image Studio turns a written prompt into an image, and Video Studio produces short clips the same way. Both run server side, both are metered against the same token balance as chat, and generations are tied to your account so you can find them again later.
Building on top: API and MCP
For automation, create a personal key on the API keys page. The secret is shown once at creation and only a hash is stored afterwards, so keep your copy somewhere safe and rotate it if you are unsure. Requests are authenticated, rate limited and charged exactly like chat — see the API documentation for endpoints and examples.
If you would rather drive WormGPT from an assistant you already use, the MCP server connects Claude, ChatGPT, Cursor and other MCP-capable clients to your account over OAuth, with the same account scoping and the same balance.
Where to go next
Open chat to try a prompt, read the acceptable use policy before you run security work, and reach us through support if something does not behave the way this guide describes.