I Found a Way To Use AI Agents Like Codex Completely For FREE

Summary

The video demonstrates Free Buff, a desktop platform that lets users run AI coding agents without paying by card. Instead, the service is supported by advertisements shown while the agent works. The presenter says the service does not require a credit card or a ChatGPT subscription, and states that he is not sponsored by or affiliated with Free Buff.

How It Works and What It Offers

• Users download the desktop app, sign in with Google, and create a project.
• Sessions are paid for with “free bucks” from a daily allowance. The allowance replenishes at midnight Pacific Time, and an hour of model use is charged at the start of a session.
• The app offers several models at different prices, including models the presenter calls GPT-5.6 Luna and GLM-5.3 Flash. He describes some options as capable enough for coding, while noting that performance varies.
• The interface includes code changes and diffs, project files, previews, and a terminal, along with build and planning modes. The presenter also points out project settings, support for an agents.md file, and an iOS option for accessing projects from a phone.
• Ads appear in the workspace, including while the agent is processing. The presenter considers them noticeable but not overly disruptive in this demonstration.

Demonstrations and Impressions

• The presenter asks an agent to make a personal one-page website. He revises the request after finding the initial result underwhelming. The final site works, but he does not consider it especially remarkable.
• He then uses GLM-5.3 Flash to make a simple stickman game. The game runs, though he finds it fairly basic and not particularly entertaining.
• He says the experience is broadly similar to using other coding agents and finds the free access useful, especially for people who cannot afford paid tools. During the test, he also notes that his computer gets warm and noisy.
• His main argument is that ad-supported access could make capable AI coding tools available to more people, even if the results and models do not match the strongest paid options.

Main speaker/source: Nick Saraev, presenting and testing the Free Buff platform.