Replit update sparks developers’ dissatisfaction over pricing

If you are a Replit user, you are most likely paying a much higher fee for using the vibe coding tool.

Several users of Replit took to Reddit this week to express their dissatisfaction over pricing changes, complaining that the tool is consuming more credits than before, specifically after the company let loose its Agent 3 update in the second week of this month.

Developers using vibe coding assistants, notably Cursor, Kiro, and Claude Code, have expressed dissatisfaction over rising costs as providers of these tools changed their pricing policy over the last few months.

The complaints from Replit users range from burning through a third of their monthly budget in one night to the upgraded agent in the tool forcefully applying changes not requested or desired.

The reasons behind the nature of the complaints, at least in part, can be attributed to the “10x more autonomous” nature of Agent 3, which Replit also claims can generate other agents and automation, resulting in Agent 3 initiating subagents to refactor code even on minor edits, leading to cost overruns.

“In the end, the refactoring is more expensive than original creation,” Jason Andersen, principal analyst at Moor Insights, said, adding that part of the users’ dissatisfaction is based on the fact that the tool is now refactoring code that was working.

The other reason for the price rise can be attributed to Replit’s effort-based pricing, rolled out in July, which calculates cost based on how much compute resources are being used rather than how many checkpoints are being created based on a user request.

Previously, a single user request could generate multiple checkpoints (basically, mapping different actions). But now, a single user request generates a single checkpoint, although the single checkpoint could end up consuming more compute resources.

Should Replit take a leaf out of OpenAI’s playbook?

Another issue of contention for developers is that Replit is not allowing users to choose a previous version of the agent, which could eliminate some of the issues that developers are facing with Agent 3, said Paul Chada, co-founder of DoozerAI — an agentic digital worker platform.

Moreover, Chada suggests that Replit would be wise to take a leaf out of OpenAI’s playbook: Offering the option to use Agent 2 alongside Agent 3, similar to how OpenAI allows users to choose between GPT-4o and GPT-5.

Additionally, Chada pointed out that Replit could have served its users better by implementing usage caps or warnings when rolling out its new pricing policy.

Seconding Chada, Moor Insights and Strategy’s Andersen said that Replit users shouldn’t expect a pricing rollback but can look forward to getting back a part of their credits consumed post the pricing change — a move that other vibe coding tool providers have executed to quell user dissatisfaction.

Replit has not announced anything to that measure yet, but CEO Amjad Masad wrote on X that the company is aware of the issues with Agent 3 and was actively trying to fix them.

As Replit users continue to seek resolution of their issues, vibe coding tool providers, including Replit, find themselves in a tricky spot: battling the rising costs of providing large language models to users who promise breakthrough productivity, while also responding to investors’ requests to stop subsidizing user capture with lower pricing.

California-headquartered Replit, which counts Andreessen Horowitz, Khosla Ventures, South Park Commons, Craft Ventures, and Y Combinator as its investors, last week raised $250 million in funding, around the same time it released Agent 3.