Skip to main content

Your free local-first
AI coding agent.

A full IDE where AI agents plan, build, and ship in parallel. Your models, your hardware, your data. No forced subscriptions here.

Download NowGet the CLI
moderation-bot
Explorer
moderation-bot
src
filters.ts
index.ts
moderation.ts
package.json
tsconfig.json
README.md
Outline
No symbols
Editor
Bodega One

Try asking the agent

Keyboard Shortcuts

New SessionCtrlAltN
Open FileCtrlP
TerminalCtrl`
Save FileCtrlS
File TreeCtrlB
Close TabCtrlW
Git PanelCtrlShiftG
Chat ModeCtrl1
ChangesCtrlShiftD
Code ModeCtrl2

Start typing or open a file

Agent
Qwen2.5 Coder 7BNew Session
moderation-botChanges happen hereForeground

Agent is ready

Type a message to start coding

AI-generated code should always be reviewed. AI can make mistakes.

Context: 12%
Interactive demo. Click anywhere to try it.

10+ providers. Your key. Your choice.

Ollama·LM Studio·OpenAI·Anthropic·Groq·Together AI·OpenRouter·vLLM·llama.cpp·LocalAI·KoboldCpp·GPT4All·MLX·Jan·Ollama·LM Studio·OpenAI·Anthropic·Groq·Together AI·OpenRouter·vLLM·llama.cpp·LocalAI·KoboldCpp·GPT4All·MLX·Jan·Ollama·LM Studio·OpenAI·Anthropic·Groq·Together AI·OpenRouter·vLLM·llama.cpp·LocalAI·KoboldCpp·GPT4All·MLX·Jan·Ollama·LM Studio·OpenAI·Anthropic·Groq·Together AI·OpenRouter·vLLM·llama.cpp·LocalAI·KoboldCpp·GPT4All·MLX·Jan·

Parallel agents. Any model. On autopilot.

The full IDE, not a chat box. Here's what the agent actually does once you hand it the work.

Bodega One — Fleet
Fleet2 running · 1 action
Add search to the settings panelActiveRunning
moderation-botqwen2.5-coder:7bsettings-search2m ago
Iter 4/25— edit: SettingsPanel.tsxLast active 3s ago
Migrate config to zod schemasRunning
moderation-botqwen2.5-coder:7bzod-config8m ago
Iter 17/25— running toolLast active just now
Fix flaky websocket testApproval needed
moderation-botdeepseek-coder-v2:16bflaky-ws-test14m ago

Fleet

A fleet of agents, in parallel

Hand off a task and it runs in the background on its own git worktree while you keep coding. Spin up several at once, each fully isolated. Watch iterations land live, then review the diff and apply with a single click. Nothing touches your working branch until you say so.

Learn more
Bodega One — Loops
Bodega Loops
Nightly doc sync
Nightly at 01:30 · next in 6h 12m
applied7/10/2026, 1:30:04 AM48s3 files8 iterQEL 92
parked7/9/2026, 1:30:02 AM1m5 files11 iterQEL 71
Weekly dead-code sweep
Weekly (Sun 03:00) · next in 2 days
Test backfill
every 4 hours

A loop runs an agent task on a schedule — every run lands as a QEL-scored, reviewable diff.

Loops

Put the busywork on autopilot

Schedule the recurring work: doc syncs, dead-code sweeps, dependency bumps, test backfills. Each loop runs on its own and lands as a reviewable, QEL-scored diff waiting for your approval. You decide what ships, and nothing merges without proof it passed.

Learn more
Bodega One — Models
Connected to Ollama — 3 models installed
High-end — 24GB

Favorites (1)

Qwen2.5 Coder 7BDefaultCode

Coding (1)

DeepSeek Coder V2 16BSmart

Fast (1)

Llama 3.2 3BFast
Local models run on your hardware — zero API costs, zero data leaves your machine

Bring Your Own LLM

Any model. Local or cloud.

Bring your own model. Run Qwen, DeepSeek, or Llama locally on your own hardware at zero cost, or drop in a cloud key from any of 10+ providers when you want more horsepower. Switch models mid-session without losing context, and air-gap cleanly so nothing ever leaves your machine.

Learn more Explore BYOLLM
Bodega One — Help & Docs

Help & Docs

Guides for every feature — chat, code mode, models, fleet, the Map, and more.

Getting Started

Bodega One is a local-first AI desktop IDE. It runs entirely on your machine — no cloud account required. This section covers first launch, connecting a model, and the core concepts you need before anything else makes sense.

What Bodega One Is

Two modes share a single window:
  • Chat mode (Ctrl+1) — conversational AI. Questions, research, quick tasks. No file tree, no AI panels.
  • Code mode (Ctrl+2) — full IDE. Monaco editor, file tree, four AI panels (Agent, Research, Debug, Advisor), and a terminal.

Connecting an LLM — Your Options

How-to
Bodega speaks to 24 provider presets. The easiest local path is Ollama — install it, pull a model, and Bodega detects it automatically. Cloud keys (Anthropic, OpenAI, and more) can sit alongside local models; pick per session.

Picking a Model That Fits Your Hardware

Tip
Bodega scores every model against your detected VRAM. Look for the green fit dot in the model picker — as a rule of thumb, a Q4_K_M quant needs roughly its parameter count in GB of VRAM.

Shortcuts

Switch to Chat modeCtrl+1
Switch to Code modeCtrl+2
Open SettingsCtrl+,
Help and keyboard shortcuts overlayCtrl+/

Built-in Docs

The manual, built in

The full manual ships inside the app, from your first launch to air-gapped setups and provider configs. It works completely offline and always matches the version you are running, so you are never stuck reading docs for a build you do not have.

Learn more Browse the docs

What else is in the box

No demo needed, no add-ons. Every one of these ships standard.

01

Quality Enforcement Layer

Most AI agents write code and move on. Bodega One Code doesn't. Every change runs through a 5-step verification loop before it's considered done, and if it fails the agent fixes it.

  • Syntax + type checks on every write
  • Test gate execution with a hard timeout
  • A scored pass/fail after every completed task
Learn more
02

Air-Gap Mode

Cut the cable entirely. Air-gap mode runs nine enforcement layers that block every outbound path, so no telemetry, model calls, or background pings ever leave the machine.

  • Nine network enforcement layers
  • Works fully offline with any local model
  • No exceptions, no workarounds
Learn more
03

Custom Agents

Build your own agents, not just prompts. A profile pins a system prompt, an optional model, a tool allowlist, a read-only flag, and an iteration cap, enforced at the execution gate.

  • Allowlist enforced at the gate: agents narrow, never widen
  • Read-only stays read-only, whatever the model tries
  • One-click starters: Code Reviewer, Test Runner, Doc Checker
Learn more
04

Bodega Map

A live, interactive graph of your project. Ask questions about your codebase in plain language and the agent navigates by the map instead of guessing.

  • Interactive file-graph of the whole project
  • Ask questions in natural language, answered by your LLM
  • Lives inside Code mode
Learn more
05

Mixture of Agents

An optional Mixture-of-Agents engine: several reference models answer your prompt in parallel, then an aggregator model reads every draft and writes the single best reply.

  • Multiple reference models answer in parallel
  • An aggregator model synthesizes the final answer
  • Optional, toggled per message
Learn more

Frequently
Asked Questions

Everything you need to know before getting started.

  • Does Bodega One Code include an LLM or AI model?+

    No. Bodega One Code is bring-your-own-LLM. We don't bundle or host any model. Run local models like Ollama, LM Studio, etc. for full privacy. Or plug in your own API keys for Claude, GPT, Groq, or any of our 10+ supported providers. You pay those providers directly at their rates. We never charge markup or touch your usage.

  • What's the difference between Personal and Pro?+

    Personal is the full product, free, for personal use on 1 machine. Every feature is included. Pro (TBD one-time, at full release) adds exactly three things: commercial use rights, unlimited workspaces, and a second machine. Pro is not on sale yet, it arrives at full release. Right now, in beta, everyone gets full access free, commercial use included.

  • Can I really run this with no internet connection?+

    Yes. If you're using local models (Ollama, LM Studio, llama.cpp), Bodega One Code runs completely offline. Air-gap mode removes web search tools entirely. Cloud API support is optional.

  • How is this different from Cursor or GitHub Copilot?+

    Cursor is subscription-only, cloud-dependent, and locks you into their model choices. GitHub Copilot is the same. Monthly fee, cloud-only, Microsoft's ecosystem. Bodega One Code is free for personal use, runs locally by default, and lets you use any LLM you want. You own it. You don't rent it.

  • Is this a finished product or a beta?+

    It's an open beta, live now, free, and open to everyone. No waitlist, no password. Download and run. Because it's a beta, expect a few rough edges, and tell us when you hit one. Full launch is coming later this year, and that's when the paid Pro tier goes on sale.

  • Can I buy the Pro tier yet?+

    Not yet. Pro (commercial use, 2 machines, TBD one-time) goes on sale at full release. You do not need it during the beta, the beta is free for everyone, commercial use included.

Try Bodega One Code.

Free for everyone in the open beta. Windows, macOS, and Linux. No account, no waitlist.

Download Now

Stay in the loop.

Product updates, build-in-public notes, and the day full launch lands.