Looking at the current stage of AI powered code editors like Cursor, Copilot, these services charge and hefty fee for providing access to premium models, even if the user might use those models for very small fraction of problem solving.

Most of the bugs can easily be solved with much cheaper models, but users are bound to pay a flat amount since they might occasionally need the over-powered AI model to just solve a linting error.

Even with the subscription, you are not guaranteed to use up the whole context window and might be missing out on a lot even when paying in full, since there is no proper proof it it.

And buying every other service separately is out of question since, it cannot manage the same context and on the other hand its even costlier

To solve this issue we have introduced the following :

• Your wallet as API keys, no managing separate keys for services, just load with USDC and pay for services
• Granular context selection, the user can select the type of context based on their choice
• The agents decide upon the complexity and cost of solving the problem, based on that model is chosen, optimising for price and accuracy
• The user pays only for the calls they made, and there's no flat subscription
• User can decide whatever their budget to be based on how accurate the solution needs to be