An early-access agentic IDE that plans, edits files, runs tests, builds and previews real projects — and generates images and video from a chat prompt. Powered by open models on your hardware, with our own brutus-brain built in. Review agent changes before trusting them.
Actual screenshot of the build you can download — agent panel, file explorer, PowerShell terminal and live preview, all in one native window.
▶ Watch it work — 5s real capture
BrutusAI doesn't just talk — it plans, writes files, runs tests, spins up servers and shows you the result live.
Plans tasks, calls tools, verifies its own output and keeps going until the job is done.
Starts dev servers, previews your site in-app — then visits it with real browser automation to verify what it built.
Use local Ollama models offline, or configure your own OpenAI-compatible provider credentials. Cloud availability, free quotas and charges depend on the provider; Discord sign-in does not supply provider credentials.
Generates images (SDXL-Lightning) and video (local text-to-video) straight from a chat prompt — fully offline, no API keys, no per-image fees.
Our own fine-tuned model, trained in-house on real agent tasks — 250 chats free on every install. It remembers techniques it learns and stops asking twice.
Ask how it feels — it reads live GPU, RAM and uptime telemetry and answers from real numbers. When the hardware's struggling, it tells you.
Microsoft neural TTS — natural spoken replies, five voices, completely free and keyless.
Watches its own services, restarts what's down, re-verifies builds, and keeps the public site live — unattended.
Searches docs, fetches pages and installs frameworks when the task needs them.
Seven modes — normal, smart, accept-edits, bypass, autonomous, plus plan & ask for read-only analysis.
Every agent edit is restorable. Sessions persist — pick up right where you left off.
Run it free on your own hardware, or let us host the brain.
Hosted plans are planned, not currently available for purchase. Prices shown are provisional. Billing setup and independent tenant isolation verification are required before launch. The local beta is available now.
A Windows PC, Ollama, and one model pull
(ollama pull qwen2.5-coder:7b). After that BrutusAI runs fully offline — no API keys, no token bills.
An access key to our hosted model server. Your laptop stops running the model — everything offloads to our infrastructure, so it's faster and works on machines with no GPU.
Our own fine-tuned model, trained in-house on real agent tasks — tool calls, coding, browser use, media generation. Your first 250 chats on it are free; Pro removes the cap. Every other model stays free and unmetered forever.
Same workflow shape — agent plans, edits files, runs commands, previews results — but powered by open models you own. Honest take: a local 7B model won't match Devin's proprietary model on huge tasks, but for everyday builds and edits it delivers, and it costs nothing to run.
Yes — run the server on any machine, hand out access keys to your team. Keys are per-user, revocable, and the free tier is rate-limited out of the box.
Deliberate. We're building the community first — the IDE is free forever, Discord membership is free, and launch pricing is set low enough that upgrading is a no-brainer. Early members keep their price locked forever, even as it goes up for everyone else.