Bodega One Code is a local-first AI coding IDE, positioned as "a complete IDE, not a chat box." It integrates AI agents, AI chat, and a full development environment, allowing agents to plan, build, and deliver code in parallel. Its core philosophy is "your models, your hardware, your data," with no forced subscription.
[Main Capabilities]
- Parallel agents: After handing a task to an agent, it runs in the background in an independent git worktree. You can set up multiple agents to work simultaneously, each fully isolated; you can observe iteration results in real time, then review the diff and apply it with one click. Your working branch is not modified until you confirm.
- Automated recurring tasks: You can schedule repetitive work such as documentation sync, dead code cleanup, dependency upgrades, and test completion. Each cycle runs independently, producing a reviewable, QEL-scored diff awaiting approval, and it must have proof of passing before merging.
- Bring your own LLM: Supports more than 10 LLM providers. You can run models such as Qwen, DeepSeek, and Llama on local hardware for zero cost, or connect cloud API keys for stronger compute. You can switch models mid-session without losing context, and air-gap mode is supported.
- Five-step verification loop: Every write undergoes syntax and type checking, a test gate with hard timeouts is executed, and a scored pass/fail result is given after task completion; if it fails, the agent fixes it itself.
- Air-gap mode: Nine layers of network enforcement controls block all outbound paths, ensuring telemetry, model calls, and background requests never leave the local machine; combined with local models, it can run fully offline.
- Custom agents (Profile): You can pin system prompts, selectable models, tool allowlists, read-only flags, and iteration limits, and enforce them at the execution gate; one-click launchers such as Code Reviewer, Test Runner, and Doc Checker are provided.
- Interactive file graph: A real-time interactive map of the entire project. You can ask questions about the codebase in natural language, and agents navigate based on the graph rather than guessing blindly. This feature is located within Code mode.
- Mixture-of-Agents (optional): Multiple reference models answer the same prompt in parallel, then an aggregator model synthesizes all drafts to write the best response. Can be enabled per message.
- Built-in manual: The complete manual ships with the application, from first launch to air-gap setup and provider configuration, and can be consulted fully offline,and it always matches the currently running version.
[How to Use] Download and install from the official website, with no account and no waitlist required. After launching, you can directly enter tasks for the agent in the IDE, or open a file and start coding; select a local model or enter a cloud vendor's API key through the configuration interface, and enable air-gap mode as needed.
[Use Cases and Limitations] Suitable for developers who want free AI coding capabilities, value data localization and privacy, need offline or air-gapped environments, and want the freedom to choose models. It can be used for personal projects and paid commercial work. In terms of limitations: the product does not include any models, so you must provide your own; when using cloud APIs, you pay the provider directly; it is currently in open testing and may have rough edges, with the official release expected within the year; official mobile support has not been mentioned.
