Bodega One

Bodega One Code is now free. For everyone. Full IDE, AI chat, autonomous agent. BYOLLM, air-gap mode, all 23 tools. No waitlist, no password. It's an open beta, so expect rough edges. Download it and break it. bodegaone.ai/download

Install one-liner + docs: bodegaone.ai/download

Bodega One Code now runs in your terminal. The CLI is live. v0.1.0, first public release. Same local-first agent, in your shell. TUI or fully headless. Your machine, your model, your keys. Air-gap mode included. Free in the open beta.

Local-first AI IDE — Chat + Code, with verification built in. Download: Bodegaone.ai/download Runs great on local @Alibaba_Qwen models. #LocalLLaMA

Bodega One beta.31.6 ships Mixture-of-Agents — but local. Several models draft your answer in parallel, then an aggregator synthesizes the final reply. All on your machine. An all-local ensemble costs about the same as one normal turn. 🧵

Also in 31.6: • Preview static sites/games with no dev server • local models no longer stuck at a 4,096-token window • long Cloud Boost builds don't cut off at 5 min

Optional: gate the synthesized answer behind a verification pass before you ever see it. Off by default. Settings → Models → Mixture.

Why local changes the math: Cloud MoA = N+1 API calls per turn. Local references are ~free, so a 3-model local mixture ≈ one turn's cost. Offline? Cloud refs drop, the local ensemble keeps running. Air-gapped Mixture-of-Agents.

Pick "Mixture" in the model picker. A few reference models (@ollama / llama.cpp / cloud) run in parallel, no tools — then one aggregator reads every draft and writes the response, owning the tools. Classic MoA. Running locally.

@araseb_ I believe the perfectionist mindset within Japan is one of their greatest attributes.

@BodegaOneAI They always perfect things in the background before making a massive move

why is Japan so behind in AI they gave us robots bullet trains PlayStation and yet barely show up in the AI race what actually happened ?

One prompt, one open model, zero cloud. Qwen3-Coder 30B built this entire 3D synthwave racer, running fully local on our GPU via llama.cpp. No API key, no subscription, all offline. Open weights are getting scary good. @Alibaba_Qwen 👀 Made in Bodega One.

4/ Also: raw llama.cpp tuning flags, a contract preview before you hit send, MCP per-server tool filtering, and QEL verification fixes across more languages. Update in-app or grab it at bodegaone.ai

1/ Bodega One beta.31 is out. The automation and verification wave: gate what your AI coding agent can do, prove the work it ships, and catch regressions before they bite. Local-first. Free.

3/ Drift Radar re-proves the work Bodega already verified against your current code and tells you what regressed. Plus proof-carrying commits and a locally signed air-gap attestation. Trust, but verify.

2/ Permission profiles and lifecycle hooks. Deny tools, confine writes by path glob, and run your own lint, format, or block command before any tool call. Hooks committed in a repo never run until you approve them.

We been working hard on improving Bodega one Code & Coming soon Bodega CLI run any @ollama Model effortless, Pull any model from within the app or 1 click download a model tailored to your hardware from our model catalog hub! Open models <3

Download here👇 bodegaone.ai/download

We just launched Bodega One Code on Product Hunt. A local-first AI IDE: full editor, AI chat, and an autonomous coding agent that runs on your machine with your own LLM. Air-gap mode means nothing leaves your laptop. Free in beta, no subscription. Last week a hosted model got switched off worldwide overnight. This is the other way to do it. producthunt.com/products/bodeg…

@16vchq Bodegaone.ai

drop ur startup link