CleverBee is an open-source AI-powered research assistant designed to synthesize information from the web. It targets researchers and power users who need to automate online data gathering and analysis, providing an interactive UI for managing complex research tasks.

How It Works

CleverBee employs a multi-LLM strategy for distinct research phases: Gemini 2.5 Pro for planning and final reports, Gemini 2.5 Flash for determining next research steps, and Gemini 2.0 Flash for summarizing intermediate web content. It uses Playwright for automated web browsing and HTML extraction, cleaning the content to Markdown before processing. The system supports modular LLM clients, including Gemini, Claude, and local GGUF models via llama-cpp-python, with an integrated SQLite-based cache for performance and cost reduction.

Quick Start & Requirements

• Install via cloning the repository and running bash setup.sh.
• Start the application with bash run.sh.
• Requires Python. macOS is fully supported (Intel and Apple Silicon via Rosetta 2). Linux is supported with NVIDIA GPU optimization for local models. Windows has limited support via WSL.
• Documentation available at: https://cleverb.ee/docs

Highlighted Details

• Interactive web UI powered by Chainlit.
• Supports external tools via the Model Context Protocol (MCP).
• Integrated token tracking and cost estimation for LLM calls.
• Configurable via config.yaml.

Maintenance & Community

• No specific contributors, sponsorships, or community links (Discord/Slack, roadmap) are mentioned in the README.

Licensing & Compatibility

• Licensed under the GNU Affero General Public License, Version 3.0 (AGPLv3).
• AGPLv3 is a strong copyleft license, requiring derivative works to also be open-sourced under AGPLv3. This may restrict commercial use or linking with closed-source applications.

Limitations & Caveats

The project explicitly states it is not a replacement for deep domain expertise and that AI can still hallucinate, requiring users to verify sources. Windows support is limited to WSL.