CraftBot with Living UI

Product Introduction

1. Definition: CraftBot with Living UI is an open-source, autonomous AI agent platform designed for desktop and server deployment. It is a "Personal AI Assistant" that operates as a persistent background service, capable of interpreting complex tasks, planning multi-step actions, and executing them to achieve user-defined goals. The "Living UI" component is a novel, integrated application framework that allows the AI agent to dynamically generate, import, and iteratively evolve custom web applications (dashboards, tools, CRMs) within its own runtime environment.
2. Core Value Proposition: The product exists to transform software from a static, purchased product into a dynamic, AI-co-created environment that adapts to the user's evolving needs. Its core proposition is proactive, context-aware automation—where the AI doesn't just respond to commands but learns user preferences and initiates goal-oriented tasks—combined with a self-evolving application layer (Living UI) that the agent can directly read, write, and act upon.

Main Features

1. Multi-Provider LLM Interface: The system features a Bring-Your-Own-Key (BYOK) architecture with a unified interface supporting major LLM APIs including OpenAI (GPT), Google Gemini, Anthropic Claude, BytePlus, and local models via Ollama. This allows users to switch providers based on cost, performance, or privacy requirements without changing the agent's core functionality. It uses a context engine with KV-cache support for optimized prompt generation.
2. Proactive Agent with Semantic Memory: The agent employs a RAG-based memory system using ChromaDB for embedding storage and retrieval. It automatically distills and consolidates daily events, building a semantic memory of user interactions, preferences, and goals. This enables proactive behavior, where the agent can suggest and initiate tasks (with user approval) aligned with learned objectives, moving beyond reactive command execution.
3. Living UI Framework: This is the flagship innovation. The Living UI is a full-stack development environment inside the agent. Users can: a) Build from scratch by describing an app in natural language, prompting the agent to scaffold a data model, backend API (Python/Go/Node.js), and React frontend. b) Import existing projects from GitHub or local code, which the agent containerizes and integrates. c) Install from a community marketplace. Crucially, the agent maintains context-awareness of the UI's state (DOM, form values, API data) and can programmatically interact with it to add features, redesign views, or trigger workflows.
4. Extensible Tool & Skill Integration: Features a dual-layer extension system. The Skill Manager loads pluggable Python modules for specific capabilities (task planning, research, code review, git operations). The MCP (Model Context Protocol) Adapter converts external MCP servers into native actions, enabling integration with a vast ecosystem of tools. Pre-configured OAuth/PKCE support is included for services like Google Workspace, Slack, Notion, Zoom, LinkedIn, and Discord.
5. Cross-Platform Service Architecture: Provides a consistent experience across Windows, macOS, and Linux. The recommended craftbot.py installer handles the entire lifecycle: dependency management, background service registration (auto-start on login), process management, and desktop shortcut creation. It also offers multiple runtime modes: Browser (default React web app), TUI (terminal UI built with Textual), and CLI for lightweight interaction.

Problems Solved

1. Pain Point: The disconnect between static software and dynamic user needs. Traditional apps are fixed upon release; customization requires manual development or complex no-code platforms. CraftBot's Living UI allows software to evolve organically through conversation with an AI.
2. Pain Point: Fragmented automation workflows. Users often juggle separate automation tools (Zapier/IFFTT), chatbots, and custom scripts that lack context and cannot perform complex, multi-domain tasks. CraftBot unifies planning, execution, and tool use into a single, memory-equipped agent.
3. Target Audience: Power Users & Prosumers who need custom tools (e.g., a unique project dashboard, a tailored CRM). Developers & Engineers seeking an AI pair-programmer that can both write code and run and modify the resulting application. Small Business Owners/Startups needing adaptive internal tools without a full dev team. Researchers & Analysts requiring an assistant to manage data, workflows, and reporting across various services.
4. Use Cases: Dynamic Business Dashboard: Creating a real-time company KPI dashboard that the agent can update, analyze, and even suggest actions based on. Personalized Learning System: Building an adaptive learning app where the agent curates content, tracks progress, and modifies the curriculum. Adaptive Workflow Automation: Designing a custom procurement or content approval pipeline that the agent can monitor, execute steps within, and redesign as process rules change.

Unique Advantages

1. Differentiation: Unlike standalone AI chatbots (ChatGPT, Claude Desktop) or automation platforms, CraftBot is a persistent, stateful agent with a dedicated memory and a tightly-integrated application runtime. Competitors like AutoGPT focus on task execution but lack a native UI framework. Custom GPTs/Assistants are cloud-bound and cannot host evolving applications locally. CraftBot runs locally/on-premise, owns the full stack, and seamlessly blends agentic reasoning with application interaction.
2. Key Innovation: The bi-directional integration between the AI agent and the Living UI runtime. The agent isn't just a user of the UI; it has programmatic access to read the UI's state, manipulate its data via API, and modify its source code. This creates a closed-loop where the tool and its intelligent operator co-evolve, a significant step towards self-improving software systems.

Frequently Asked Questions (FAQ)

1. Does CraftBot with Living UI require coding knowledge? No, primary interaction is through natural language. You can describe the app you want, and the AI will handle the initial coding. However, reviewing and refining the generated code benefits from technical understanding.
2. Is my data private with CraftBot? Yes, when using local LLMs (Ollama) or BYOK with trusted cloud providers, your data, memory, and Living UI apps reside on your machine. The core platform is open-source and runs as a local service, ensuring data sovereignty.
3. How does CraftBot's "Living UI" differ from a traditional web app? A traditional web app is static; new features require manual development. A Living UI is dynamic and mutable; its features, layout, and data model can be altered through natural language commands to the embedded AI agent, which can directly rewrite the application's code and logic.
4. Can I connect CraftBot to my company's internal systems (e.g., Slack, Google Workspace, Jira)? Yes, via its built-in OAuth integrations for major platforms and its MCP Adapter, which can connect to any tool exposing an MCP server, including custom internal tools, databases, or APIs.
5. What are the system requirements to run CraftBot? It requires Python 3.10+, Git, and moderate system resources. The craftbot.py installer automates setup. For the Browser UI, Node.js 18+ is needed. It runs on Windows, macOS, and Linux, with full Docker container support for isolated deployment.