Manus shifts to credit-based pricing

Charging in March 2025 turned Manus from a viral demo into a meter on expensive AI work. The product does not just answer a question, it opens browsers, reads pages, runs code, and calls outside tools, so each task burns real compute and API spend. A paid, credit based plan let Manus convert a 2 million person waitlist into revenue while matching price more closely to underlying cost than a flat chat subscription would.1 2 3

• The jump from invite only beta to paid plans came right after Manus packaged browser control, deep research, and 29 integrations into one consumer agent. That made monetization possible because users were buying finished work, like reports, forms, and slide decks, not just model access.1 2
• Its pricing logic is really infrastructure pricing disguised as SaaS. Tasks can consume hundreds of credits because Manus is paying for model tokens, virtual machine runtime, and third party API calls. That is why a credit wallet fits better than unlimited use.2 4
• A close comparable is Perplexity, which also started with a simple subscription and then added credits as agent workflows got heavier. In both cases, the move signals that deep research agents are less like search boxes and more like cloud software with variable unit costs.3 5

The next step is more price segmentation, more enterprise packaging, and tighter controls on usage. As agents take on longer tasks and connect to more systems, the winners will be the companies that turn variable compute into predictable customer spend without choking off the heavy users driving most revenue.4 5