openaccountants/openaccountants

What it solves

AI models often hallucinate or rely on outdated training data when answering tax questions, and they cannot provide professional accountability. OpenAccountants provides a library of tax guides that are reviewed by licensed accountants, allowing AIs to cite specific, verified sources and name the professional who reviewed the information.

How it works

The project maintains a repository of tax guides across hundreds of jurisdictions. These guides exist in two states: accountant-reviewed (verified by a named professional) or source-cited drafts. The system uses a Model Context Protocol (MCP) server to connect these guides to AI clients like Claude, ChatGPT, and Cursor. To maintain accuracy, the project employs automated checks for mathematical consistency, source monitoring for page changes, and a nightly exam to ensure AIs are not improvising answers.

Who it’s for

• AI Users: People seeking reliable, cited tax information from their AI assistants.
• Accountants: Licensed professionals who want to contribute their expertise to AI knowledge bases and receive public attribution.
• Developers: Those integrating verified tax knowledge into AI applications via MCP or manual file imports.

Highlights

• Professional Accountability: Every reviewed guide includes the name and credentials of the licensed accountant who verified it.
• Citable Sources: Moves AI answers from training-data guesses to cited, current guides.
• MCP Integration: Easy connection to popular AI clients via a hosted connector or a self-hosted Python MCP server.
• Automated Quality Control: Nightly syncs that check for math errors, stale sources, and AI improvisation.