arman-jalili/rigorix-oss

Most coding agents improvise step by step, which makes them hard to audit when something goes wrong. Rigorix flips that by forcing the plan into a fixed graph you can inspect and approve before a single line of code runs. Teams that need CI grade repeatability and permission controls around AI code generation are the obvious audience, not solo hackers who just want speed. It is early and tiny right now, but the separation of planning from execution is a genuinely different bet than the usual agent loop design.