Rigour is not a linter. It's a three-layer immune system for AI coding agents — combining deterministic AST gates, a model trained on real AI failures via RLAIF, and a Bayesian local brain that learns your codebase patterns over time

AI agents are getting insanely capable. The real problem now isn’t generating code — it’s controlling what agents can do, proving what they changed, and making each run smarter than the last. Rigour is the open-source control plane for AI coding agents: context, policy, deterministic verification, MCP governance, signed evidence, Fix Packets, memory, and learning. Let agents move fast. Keep engineering under control.

I built Rigour because AI made me faster — but not necessarily more confident. As coding agents became capable of changing larger parts of a codebase, I kept running into the same problem: the smarter the agent became, the harder it was to understand everything it changed, verify every decision, and carry the lessons from one session into the next. I didn’t want another AI wrapper or another LLM judging an LLM. I wanted a deterministic layer around the agent — one that understands the codebase, limits blast radius, verifies observable work, produces evidence, and learns only from outcomes that were actually proven. That became Rigour. Today it connects context → agent actions → policy → verification → evidence → learning, and can govern high-impact MCP actions through a trusted gateway. The idea is simple: AI agents will keep getting more powerful. The engineering system around them needs to get stronger too. Rigour is my attempt to build that missing layer — open source and local-first.

Rigour

The immune system for AI coding agents

The immune system for AI coding agents

Rigour Labs

Previous Rigour Launches

Previous Rigour Launches