As businesses continue to embrace artificial intelligence, many are discovering an unexpected challenge: rapidly increasing AI usage costs. A new open-source tool known as the Caveman AI plugin is emerging as a practical solution, helping companies significantly reduce their AI expenses by making chatbots communicate with fewer words while maintaining the same technical accuracy.
The plugin transforms the responses of AI coding assistants such as Claude Code, OpenAI Codex, and Google’s Gemini into short, direct replies, eliminating unnecessary conversational language that consumes valuable AI tokens.
According to a report by 404 Media, the tool was developed in response to the growing financial burden associated with enterprise AI adoption, where companies are spending millions on AI-generated responses.
Why AI Tokens Matter
Large language models charge users based on the number of tokens processed. Tokens include both user prompts and the AI’s responses, meaning lengthy explanations and conversational phrases increase operating costs.
Julius Brussee, the creator of Caveman, explained that he built the plugin after noticing that a large portion of his AI spending came from unnecessary language such as greetings, polite expressions, transitions, and lengthy explanations.
Rather than changing the technical output, Caveman simply removes the extra wording surrounding it.
Instead of detailed conversational responses, the AI communicates using brief, straightforward statements while preserving code, commands, URLs, file paths, function names, and other technical information.
Companies Are Looking for Ways to Reduce AI Spending
The growing interest in Caveman reflects a broader effort by businesses to control AI expenses.
As reported by 404 Media, several organisations are actively introducing policies to reduce AI token consumption.
One example is Legrand, a global electrical and digital infrastructure company. An internal memo reportedly advised employees to use AI more efficiently by selecting appropriate models, avoiding unnecessary high-reasoning settings, and using the Caveman plugin to reduce output consumption without affecting code quality.
The report also notes that developers at companies including OpenAI, Nvidia, GitHub, and DEPT have either adopted or experimented with the tool.
Significant Token Savings
One of Caveman’s biggest advantages is its ability to reduce output tokens dramatically.
According to its creator, the plugin can lower output token usage by approximately 65% to 75% compared to standard AI responses.
Users can even monitor how many tokens have been saved over time.
During testing by 404 Media, the plugin reportedly reduced token usage by about 5,800 tokens, representing roughly a 65% reduction.
The plugin offers several response styles, including:
- Lite
- Full (default)
- Ultra
- Wenyan, which translates responses into classical Chinese
The objective remains the same across all modes: minimise unnecessary language while preserving essential technical accuracy.
OpenAI Engineer Contributed to the Project
The project has also attracted attention from major technology companies.
GitHub records show that Shayne Sweeney, Director of Engineering at OpenAI, contributed code to Caveman by adding support for OpenAI’s Codex platform.
The project has since expanded beyond a simple plugin and now includes a complete AI coding agent capable of generating concise responses throughout an entire development workflow.
Caveman also supports OpenClaw, another popular AI agent platform.
Rising AI Costs Are Becoming a Major Business Challenge
The popularity of Caveman highlights a wider industry trend as organisations seek better control over AI spending.
Several companies have already introduced measures to manage AI costs.
GitHub recently announced pricing changes that charge customers based on token consumption rather than a flat subscription fee.
Uber reportedly exhausted its annual AI budget within just four months, prompting restrictions on employee AI usage.
Walmart has also introduced limits on AI tool usage to better manage operational expenses.
Even OpenAI CEO Sam Altman previously joked that users saying “please” and “thank you” to AI systems collectively cost the company tens of millions of dollars in electricity and computing resources.
As businesses continue integrating artificial intelligence into daily operations, tools like Caveman demonstrate that improving efficiency is no longer just about building better AI models. It is also about using them more economically.
Source: As reported by 404 Media.
For more interesting news and articles, checkout our blog page https://www.absmiley.com/blog/