OpenThesis is an open-source, evidence-first research workbench for public companies. It grounds AI analysis in official filings, uses deterministic code for financial calculations, and lets specialized AI agents focus on reasoning instead of guessing facts. It supports US, A-share, and Hong Kong markets, multiple AI providers, and produces traceable research with sources you can verify.
Framer AI AgentsDesign and publish professional sites with AI
Promoted
Maker
📌
I built OpenThesis after repeatedly running into the same problem with LLM-based company research: the reasoning could be useful, but the financial facts underneath it were often much harder to trust.
Wrong reporting periods, inconsistent accounting scopes, missing values, and calculations performed by the model can all quietly affect the final conclusion.
OpenThesis takes a different approach: start with official filings, extract evidence first, handle deterministic financial calculations in code, and then let specialized AI agents focus on reasoning, synthesis, and verification.
It’s open source and still evolving. I’d especially love feedback on the evidence-first workflow, the research experience, and which markets or research features you’d like to see next.
GitHub: https://github.com/zjy1346/OpenT...