AI data governance, without an enterprise stack
"AI data governance" usually means a six-figure enterprise product. If you are a solo founder, a consultancy, or a small studio, you still need to answer the same question: which AI touched which file, and was it allowed to? Kobel is the small-team answer.
The three jobs of a governance layer
- Scope — decide up front what each AI may touch.
- Enforce — block anything outside that scope.
- Audit — know after the fact what actually happened.
Kobel does all three on a single Windows machine, without a server, without a subscription you cannot afford, and without exposing your files to yet another cloud.
Scope with five colors
Red, orange, yellow, teal, green. Each folder gets one. The color is the policy.
Enforce via MCP
Every AI assistant that speaks MCP goes through Kobel. Requests outside the policy are refused at the gateway.
Audit in one log
Every read, write, and delete is recorded with timestamp and AI identity. Export to CSV for your records.
Data residency
Nothing leaves the device. Permission checks are local. File contents never reach Kobel's infrastructure because there is no Kobel infrastructure in this path.
Download KobelFrequently asked questions
Is Kobel GDPR-friendly?
Kobel processes permission metadata locally. It is not a data processor for your files. See the privacy policy on kobel.app for details.
Can I export the audit log?
Yes, as CSV.
Does Kobel work offline?
Yes. Once installed and configured, Kobel does not require internet for permission evaluation.
How is this different from a DLP tool?
DLP tools focus on human users and email. Kobel focuses on AI agents acting on your files via MCP.
Team version?
The current release is single-machine. Multi-machine policy sync is on the roadmap.