AI generated image for representation As companies grapple with rapidly rising AI costs, some are turning to an unusual solution: making chatbots speak like cavemen. A plugin called Caveman strips away polite phrases, unnecessary explanations and conversational filler from AI responses, helping reduce token usage and lower operating costs without affecting technical accuracy.The tool is designed for coding assistants such as Claude Code, Codex and Gemini, replacing lengthy replies with short, direct responses. Instead of saying, “You’re right to push back, I was wrong,” an AI using the plugin might simply respond with something closer to “Hulk smash.”
Why companies are making AI less talkative
The move comes as businesses try to control soaring and often unpredictable AI expenses. Every word generated by an AI model consumes tokens, which directly affect usage costs.According to Caveman creator Julius Brussee, he developed the plugin after noticing that much of his spending on Claude Code came from unnecessary conversational language rather than useful technical information.The plugin removes pleasantries, hedging, transitions and other verbose text while leaving code, commands, URLs, numbers and technical details untouched.
How much can Caveman reduce AI costs
Brussee says the plugin typically cuts output tokens by 65 to 75 per cent compared with the default responses generated by large language models.In tests shared by 404 Media, Claude Code produced noticeably shorter replies. After reviewing the code, it simply asked, “Want changes to it?” and described functionality with concise statements such as “Uses official API, not scraping.”The plugin also tracks token savings. In one test, it reported saving around 5,800 tokens, equivalent to roughly 65 per cent of the output.
Companies are encouraging employees to use it
The plugin has already attracted interest from developers working at companies including OpenAI, Nvidia and GitHub, according to Brussee.An internal memo from electrical and digital infrastructure company Legrand reportedly advised employees to use the “caveman skill” as one way to reduce AI consumption after changes to the company’s billing system and usage quotas.The memo also encouraged employees to avoid using the most powerful AI models for every task and to select reasoning settings more carefully to stay within budget.
Different ‘grunt’ levels for different users
Users can choose among multiple response styles, including Lite, Full, Ultra, and Wenyan, the latter of which translates responses into classical Chinese.The project’s GitHub repository also includes a standalone coding agent called caveman-code, which claims to use around half as many tokens as OpenAI’s Codex while supporting more than 20 AI providers.GitHub records also show that Shayne Sweeney, Director of Engineering at OpenAI, contributed code adding Codex plugin support to the project.
Why AI token costs are becoming a bigger concern
The emergence of Caveman reflects a broader industry effort to rein in AI spending. GitHub has shifted customers to token-based billing instead of flat-rate subscriptions. Uber reportedly exhausted its AI budget within four months before introducing usage limits, while Walmart has also capped employee access to AI tools.Earlier, OpenAI CEO Sam Altman said that even users saying “please” and “thank you” to AI models collectively cost the company tens of millions of dollars in electricity expenses.Although the plugin began as a humorous project, its underlying purpose is serious: reducing unnecessary AI output to make large-scale deployment more affordable.As Caveman’s GitHub page puts it, “Caveman save you token, save you money.”