Open-source, AI deep research

I turned my 40-tab research habit into an open-source tool — cleverb.ee

Hey IH 👋 I’m Nick, a CEO who spends too many hours collating sources for strategy docs.

I loved what OpenAI and Google did recently with their Deep Research tools, but the results I found quite disappointing, and still contained hallucinations, and were just a bit painful/wordy to read.

I built cleverb.ee so the grunt work happens while I sip coffee.

What it does
• Crawls webpages, PDFs, Reddit threads, YouTube transcripts, PubMed, etc.
• Uses a tiered Gemini stack (2.5 Pro → analysis, 2.5 Flash → agent steps, 2.0 Flash → summaries).
• Outputs a balanced, fully cited Markdown report with token-cost breakdown.

It supports MCP tools, and cloud LLM or Local LLM (e.g. DeepSeek, Lllama) providers. It actively seeks out unbiased, balanced information, contests assumptions, and cites its sources.

Why open source?
Knowledge should be free, open, transparent, shared - it benefits everyone.
I wanted full transparency on every agent step and a place for others to bolt on new data sources. It’s AGPL-3.0; contributions welcome.

Quick start

git clone https://github.com/SureScaleAI/cleverbee
cd cleverbee && bash setup.sh