Anterior’s symbolic-reasoning separates gathering facts from executing deterministic rules. It compiles .srg rules, requests information from an LLM, database or person, and records a decision trace. The Apache 2.0 project supports Python 3.12 or later and stdio MCP. Although developed around medical policies, it also proposes uses in other rule-heavy domains. This is a pre-1.0 research tool whose API can change. Consistent execution does not guarantee correct inputs or rules, so consequential outputs still require review.
Back to Tech
Separating LLM fact gathering from deterministic rules
Anterior's open library exposes traceable reasoning through Python and MCP.