We believe AI tooling should be transparent and verifiable.
AI agents got powerful fast. Governance didn't keep up. FORG closes that gap — an on-device control plane that secures, governs, and measures every agent action without ever touching the work itself.
Three principles, no exceptions
Metadata-only architecture
FORG is built so it never needs your prompts or code. Session timing, token counts, model names, and rule results — that's the complete surface. Privacy is the design.
Open update channel
Every agent binary is Ed25519-signed with published SHA-256 sums. The updater is fail-closed: a missing or invalid signature means the update is refused, no exceptions.
Honest numbers
We label estimates as estimates, show provenance on every stat, and never fake a zero to look better. If a number is modeled, it says est. If it's measured, it says where.
We're a small team shipping fast.
We hire people who delete more code than they add, label their estimates, and care about the 3am page. If that sounds like you, we'd like to hear from you.
See open roles →Govern your agents with us
Start your 14-day trial, or come build the control plane.