Summary

DotCraft is an open-source, self-hosted AI agent harness built on .NET, designed to integrate AI capabilities directly into software development projects. It addresses the challenge of fragmented AI agent interactions by organizing workflows around a project's folder structure, enabling a unified workspace. This allows multiple entry points—such as Desktop, CLI, TUI, chatbots, and APIs—to share a single session core, configuration, skills, and observability surface, providing a cohesive and context-aware AI experience for developers and power users. The primary benefit is a deeply integrated AI agent that understands the project's context, leading to more efficient and controlled AI-assisted development workflows.

How It Works

DotCraft employs a "project-first" approach, treating each project directory as its own workspace. This ensures that each project has a dedicated agent with its own conversations, memory, skills, plugins, and settings, fostering a nuanced understanding of the specific project context. It utilizes a unified session model where various applications, from Desktop clients to chatbots, share the same execution engine, maintaining conversational continuity across different interfaces. The system supports multiple AI model providers, including OpenAI and Anthropic protocols, and can leverage existing ChatGPT subscriptions, offering flexibility in model selection. Observability features like traces, dashboards, and sandbox settings provide control and transparency over agent operations.

Quick Start & Requirements

The recommended installation is via the Desktop application, available for Windows, macOS, and Linux from GitHub Releases. Alternatively, CLI and TUI binaries can be installed via shell scripts. Building from source requires the .NET 10 SDK and Rust toolchain. Users must configure an AI model provider, which can be OpenAI, Anthropic, or compatible services, or utilize an existing ChatGPT subscription. The setup involves initializing a workspace by selecting a project folder and configuring the model provider, which can be done through the Desktop wizard or by editing configuration files. The official documentation and a detailed getting started guide are available online.

Highlighted Details

• Project-First Workspaces: Each project gets its own agent, ensuring deep context awareness.
• Unified Session Model: Seamlessly switch between Desktop, CLI, TUI, and chatbot interfaces while maintaining conversational continuity.
• Multi-Provider Support: Integrates with OpenAI, Anthropic, and compatible LLM providers, including native support for DeepSeek V4 and MiMo V2.5.
• ChatGPT Subscription Integration: Allows users to leverage their existing ChatGPT Plus/Team/Enterprise subscriptions without additional API costs.
• Agent Builder: Customize agents through conversational refinement.
• Teams: Facilitates multi-agent collaboration for complex tasks.
• Dreams: Background memory consolidation with user approval for memory entries.
• Interactive Tool Cards: Renders tool results as actionable UI elements within conversations.
• Cross-Channel Support: Deploy agents across various chat platforms like Telegram, WeChat, and others.

Maintenance & Community

The project is inspired by and built upon established agent frameworks like nanobot and codex. Specific community channels (e.g., Discord, Slack) or active contributor details are not explicitly highlighted in the provided README.

Licensing & Compatibility

DotCraft is licensed under the Apache License 2.0. This license is permissive and generally allows for commercial use and integration into closed-source projects.

Limitations & Caveats

Requires a .NET 10 development environment for building from source. The system relies on external LLM providers, though it offers flexibility in choosing and configuring them. While the documentation suggests a straightforward setup, the depth of integration with project files implies a learning curve for advanced customization and troubleshooting.