Microcosm turns a codebase into a substrate an AI coding agent reads to orient itself before it acts, instead of grepping around blind. It also keeps a verifiable record of what the agent actually did, not what it claims it did. The part I care about isn't the feature list, it's the discipline underneath: ground claims in evidence, calibrate before asserting, check custody before trusting a source.
Source-available, Apache-2.0. Two-minute demo, which is the fastest way to judge whether it's nonsense or not:
https://drive.google.com/file/d/1CETMi4OoYIu5N6IcIIVF7eeATFG...
The honest situation: I've been building this in a bubble and I don't know if it's useful to anyone but me, or how you get something like this seen when you have no audience. So I'm asking the people here. If you look, I'd much rather have a blunt "this is just reinventing X" or "this doesn't solve a real problem" than silence. And if you've shipped something solo and got it noticed, how did you actually do it? It's been the hardest part of the process for me and I'm jsust at my wits end about how to get anyone to even tell me how much it sucks lmao.
Thank you so much and have a good day.