Overview / Description
Kremis is a tool designed to validate claims made by large language models (LLMs) using empirical data. It ingests entity-attribute-value (EAV) signals and constructs a deterministic graph to classify each response into three categories: FACT (directly supported by the data), INFERENCE (derived from the data through logical reasoning), or UNKNOWN (not present in the dataset). Kremis operates without confidence scores, providing clear-cut classifications. Built with Apache 2.0 licensing, it is implemented in Rust for performance and reliability, ensuring ACID compliance via redb database technology. The tool offers an HTTP API, a command-line interface (CLI), and a MCP bridge for integration with models like Claude/Cursor.
Used For
AI tool for fun workflows
Pricing
Pros & Cons
Pros
• Classifies LLM responses as FACT, INFERENCE, or UNKNOWN against real data • Deterministic, graph-based — no fuzzy confidence scores • Built in Rust with ACID guarantees via redb • Offers an HTTP API, CLI, and an MCP bridge for Claude/Cursor • Open-source under the Apache 2.0 license
Cons
• Requires structured EAV data to validate claims against • Developer-oriented — not a plug-and-play consumer tool • Early, niche project with a limited ecosystem
Questions & Answers
Alternatives
Compare this tool against close alternatives in the same category, focusing on output quality, onboarding speed, and workflow fit.