Dodaios v4.2.3: Open-Source Superagent for Desktop

What changed

Dodaios, previously known as Personal Jarvis, has released version 4.2.3, introducing itself as an open-source superagent designed to run entirely on a user's desktop. The core functionality revolves around a single 'boss' agent, Dodaios, which manages and delegates tasks to other specialized agents. Users can interact with Dodaios via voice or chat, and the superagent will formulate briefs and assign work to the appropriate agents, continuing tasks even after the user disconnects. A key feature is the ability to run multiple coding agents simultaneously in separate panes, allowing for parallel task execution such as running tests and fixing identified issues. Users can also create custom agents for roles like researchers or release managers, each maintaining its own conversation history and learning from its work. Dodaios incorporates a safety system that assigns risk levels to actions, requiring user approval for any changes that could affect the system. The interface is designed with a gamified approach, where agents and users earn experience points and rise in rank as work is completed, with agents visualized as crew members on a ship.

Dodaios operates without its own account, paid tiers, or cloud infrastructure. It integrates with existing AI subscriptions such as Claude or ChatGPT plans, or can utilize local models through platforms like Ollama. The project is licensed under Apache 2.0. Installation is facilitated through simple scripts for Windows, macOS, and Linux, with no requirement for Python, Git, or a GPU. Updates are handled by re-running the installer. The system architecture shows Dodaios at the center, planning and delegating to coding panes, specialized agents, browser/app interfaces, and a local memory/wiki.

Dodaios supports a wide range of coding agents including Claude Code, Codex, OpenCode, Kimi Code, GLM, Grok Build, Antigravity, Cursor CLI, and DeepSeek Harness. It is compatible with models from OpenAI, Claude, Gemini, OpenRouter, NVIDIA, and Ollama, or any OpenAI-compatible server. Agents can also run on Hermes and OpenClaw. The application is available for Windows, macOS (Apple Silicon and Intel), Linux, and servers without a screen. Additional features include offline voice command capabilities, an 'Artifacts' wall for storing agent-generated reports, and a marketplace for community-contributed agents, skills, and plugins.

Why it matters for builders

For AI builders, Dodaios offers a unified platform to orchestrate and manage diverse AI agents locally. This reduces the complexity of integrating multiple AI tools and services, providing a consistent interface for delegation and monitoring. The ability to run agents on personal hardware enhances data privacy and control, while the support for various model providers and coding agents allows builders to leverage their preferred tools and existing investments. The gamified interface and agent ranking system can also foster a more engaging development experience.

Practical impact

Developers can install Dodaios using straightforward command-line scripts for their respective operating systems. They can then integrate their existing AI subscriptions or local models to start building custom agent workflows. The project encourages contributions, with clear guidelines for first-time contributors and security reports. Builders can experiment with running multiple coding agents in parallel for complex development tasks or create specialized agents for repetitive workflows, leveraging the marketplace for community-developed extensions.

Caveats and source limits

The source mentions that macOS disk images are not yet notarized, requiring manual user approval on first launch. Windows installers are pending approval from the SignPath Foundation for code signing. While the project is open-source under Apache 2.0, specific details regarding performance benchmarks or independent evaluations of its agent orchestration capabilities are not provided in the excerpt. The release date of v4.2.3 is not explicitly stated, though the excerpt mentions "fresh release."