CraftBot

Product Introduction

1. Definition: CraftBot is an open-source, self-hosted proactive AI assistant and agentic framework designed to run locally on a user's machine. Classified as an autonomous agentic loop system, it leverages Large Language Models (LLMs) to interpret natural language instructions, architect multi-step plans, and execute complex workflows 24/7.

2. Core Value Proposition: CraftBot exists to bridge the gap between static LLM chat interfaces and true autonomous task execution. By utilizing a "Bring Your Own Key" (BYOK) model and a local-first architecture, it provides a secure, privacy-centric environment for task automation. Its primary value lies in its proactive nature—unlike standard AI bots that wait for prompts, CraftBot learns user habits and life goals to suggest and initiate tasks, optimizing daily productivity through persistent background execution.

Main Features

1. Bring Your Own Key (BYOK) Multi-LLM Support: CraftBot features a unified LLM interface that abstracts the underlying provider. It supports proprietary models from OpenAI, Google Gemini, Anthropic Claude, and BytePlus, as well as local inference via Ollama. This flexibility allows users to swap models based on cost, latency, or reasoning capabilities without changing their core workflow.

2. RAG-Driven Memory System: The platform incorporates a sophisticated Memory Manager utilizing ChromaDB for vector-based semantic storage. It employs a "midnight distillation" process where the agent reviews and consolidates the day’s events into long-term memory. This Retrieval-Augmented Generation (RAG) approach ensures the agent maintains context over months of operation, learning user preferences and historical project data.

3. Model Context Protocol (MCP) and Skill Framework: CraftBot implements the Model Context Protocol (MCP) to extend its capabilities via external tools and services. Its extensible skill framework includes built-in modules for research, code review, git operations, and task planning. This modularity allows developers to inject custom Python-based skills directly into the agent’s context, enabling it to interact with specific local or cloud APIs.

4. Proactive Agentic Loop: Beyond reactive task execution, CraftBot utilizes an Action Router and State Manager to maintain a continuous agentic loop. It monitors user objectives and proactively proposes plans (subject to user approval). This feature transforms the AI from a tool into a collaborator that manages to-do lists, identifies bottlenecks, and initiates research tasks based on life goals.

5. Multi-Interface Accessibility: CraftBot supports four distinct interaction modes: a modern Node.js-based Browser UI for ease of use, a Terminal User Interface (TUI) built with the Textual framework for lightweight interaction, a standard CLI for scripting, and an experimental GUI mode. The GUI mode utilizes OmniParser and Docker-based screen automation to allow the agent to "see" and interact with desktop applications visually.

Problems Solved

1. Privacy and Data Sovereignty: Many enterprise AI assistants require uploading sensitive data to third-party servers. CraftBot solves this by being self-hosted and open-source (MIT License), ensuring that interaction history, local files, and OAuth credentials remain on the user's local hardware or private container.

2. Fragmented Digital Workflows: Users often struggle to coordinate data across Slack, Notion, Google Workspace, and Zoom. CraftBot acts as a central orchestration layer, using integrated OAuth/PKCE authentication to execute tasks across these platforms autonomously, such as summarizing meetings and updating project boards.

3. Target Audience:

• Software Engineers: Seeking to automate code reviews, git operations, and documentation.
• Productivity Enthusiasts: Users looking for a "Second Brain" that can execute tasks rather than just store notes.
• Data-Sensitive Professionals: Individuals in healthcare, finance, or law who require AI assistance without compromising data privacy.
• AI Researchers: Developers testing agentic workflows and Model Context Protocol (MCP) implementations.

4. Use Cases: CraftBot is essential for automated daily standup preparation, proactive life goal tracking (e.g., suggesting a workout or research topic based on a calendar), complex cross-platform data migration, and persistent monitoring of digital environments for specific triggers or events.

Unique Advantages

1. Proactivity vs. Reactivity: While most AI agents (like AutoGPT or standard GPTs) are strictly reactive, CraftBot’s core architecture is built for proactive behavior. It anticipates needs based on the "Memory System" and "State Manager," moving the user from a "commander" to a "supervisor" role.

2. Native Desktop Automation: The integration of OmniParser in GUI mode allows CraftBot to perform tasks that don't have APIs. By visually interpreting UI elements, it can automate legacy software or desktop-only applications, a significant leap over web-only agents.

3. Hybrid Deployment: CraftBot offers the versatility of running as a background service on Windows/Linux or within a Docker container. This allows it to function as a persistent "digital employee" that starts on login and maintains state across system reboots.

Frequently Asked Questions (FAQ)

1. Is CraftBot completely free to use? Yes, CraftBot is open-source under the MIT license. However, because it follows a "Bring Your Own Key" (BYOK) model, you are responsible for any API costs incurred from providers like OpenAI or Anthropic. You can use it entirely for free by connecting it to a local Ollama instance.

2. Does CraftBot require high-end hardware? For the standard Browser, TUI, and CLI modes, CraftBot is lightweight and runs on most modern systems. The experimental GUI mode, which involves UI element detection (OmniParser) and screen automation, typically requires a GPU and approximately 4GB of additional space for model weights.

3. How does CraftBot handle my login credentials for Slack and Google? CraftBot uses OAuth 2.0 and PKCE for secure authentication. Credentials can be stored in a local .env file or managed via embedded credentials in release builds. Since the project is self-hosted, your tokens never leave your machine; they are only used to facilitate direct communication between your local agent and the service provider’s API.

4. Can I extend CraftBot's capabilities with my own scripts? Absolutely. CraftBot is designed with a pluggable Skill Manager. You can add custom Python scripts to the skills directory, which the agent will automatically detect and incorporate into its action library, allowing it to perform specialized tasks unique to your workflow.