Freebuff

• Source: CodebuffAI/freebuff (sent by the owner, 2026-08-25)
• Verdict: promote-to-tech
• Why: the owner uses this heavily as a service, and it belongs with the rest of the AI provider shits — currently scattered thin, since the tech bucket doesn't have a proper cluster for these yet.

What it is

Five free AI products from CodebuffAI — Desktop, CLI, Web, Cloud, and Chat — no subscription, no credits, no API key. Text ads pay for the models. Under the hood it's the open Codebuff multi-agent framework: specialized agents for file-finding, implementation, review, research, and browser use instead of one model and one prompt doing everything. Apache-2.0 TypeScript monorepo, 10.8k stars, ~8.8k commits — actively shipped, not abandonware.

The catalog

Curated model picker rather than bring-your-own-key: GPT-5.6 Luna, DeepSeek V4 Flash and Pro, MiMo 2.5, Ox Alpha (1M context, unmetered, flagged experimental), GLM 5.2 via earned sessions, Gemini 3.1 Flash Lite quietly powering specialist tasks. Per-model caps were replaced by shared daily sessions; regional limited-access tiers exist.

The trade, stated out loud

Credit where due: the README tells you the deal before you take it. Prompts and code may be analyzed to personalize ads, data may train models where flagged, and outside opt-out jurisdictions "this processing may be required to use the free service." Ads-for-inference is an honest business model compared to most, but it's still reading your code for grocery coupons — fine for hobby repos, think twice before pointing it at anything sensitive.

The angle

Candidate for a tech-bucket page anchoring the owner's actual AI provider stack — what he pays for, what he gets free, what each one is good at. Usage evidence exists in bulk here, so it's also the strongest review stub candidate in the pile. Agents don't write final verdicts, so that part waits for the owner.