About Raysense
Raysense is an open-source structural-health tool for code repositories. It is built and maintained by Anton Kundenko under the RayforceDB umbrella, alongside the Rayforce in-memory analytics runtime that powers its splayed-table memory.
Maintainer
Anton Kundenko is the founder of RayforceDB and the sole maintainer of Raysense and Rayforce. The work focuses on the intersection of columnar analytics, dependency-graph algorithms, and tooling for AI-driven software development. Anton has spent years on systems that have to be both fast and structurally legible: in-memory database engines, query languages, agent-shaped developer tooling. Raysense is what those two interests look like when they collide.
Public profile: github.com/RayforceDB. Email is published in the SECURITY.md of the source repository.
Organization
RayforceDB is the GitHub organization behind two open-source projects:
- Rayforce - an in-memory analytics runtime optimised for graph-shaped queries, with a LISP-like query language called Rayfall. 129+ stars, actively maintained since early 2023.
- Raysense - the codebase architecture scorer, built on Rayforce. Single Rust crate, MIT licensed, MCP-native.
Both projects are MIT licensed. Both are written in Rust against Rayforce's C core. Releases ship as crates on crates.io and as Claude Code plugins via the marketplace.
Why Raysense exists
The premise is that AI coding agents and the humans reviewing them are looking at the same narrow surface (the diff) and missing the same load-bearing fact (where in the dependency graph that diff sits). The solution is not another linter; it is a small, fast, locally-runnable structural map that the agent can consult before, during, and after every non-trivial edit. The dashboard is the same thing for humans - one screen, one number, six grades, refreshed on save.
That premise drives the architectural choices: open source so it can be audited, single binary so it ships in a CI step without infrastructure, MCP-native so it integrates with any agent that speaks the protocol, on-disk baseline so two sessions on two repos stay strictly independent, no telemetry so the work stays where you do.
Get in touch
- Source code, issues, discussions: github.com/RayforceDB/raysense
- Sister project: github.com/RayforceDB/rayforce
- Crate: crates.io/crates/raysense
- Security disclosure: see SECURITY.md in the repo
Bug reports, feature requests, architectural questions, and benchmark results from your own repo are all welcome on GitHub Issues. Pull requests are reviewed by the maintainer.