Skip to content

Kanuki

Verified permission boundaries for AI agents

  • Wilmington, New Castle County, DE
  • Founded 2026
  • Cybersecurity
  • kanuki.io

Raised to date

$35K

of $95K offered across 1 Form D filing

Still open
$60K
Investors
5
Last filed
Mar 30, 2026

About

Kanuki is a security company that builds permission management for AI agents. Instead of enumerating what an agent is allowed to do, it has organizations define explicit no-go boundaries - actions an agent must never be able to take, such as deleting a database, initiating a payment, escalating privileges or reaching sensitive data - and then formally verifies those boundaries cannot be crossed, including via indirect system paths. Access is enforced within the verified boundaries and every action is logged for audit. The company was founded by Lisa Akselrod and raised an angel round in early 2026.

Uses formal verification to mathematically prove agent boundaries hold, rather than relying on allow-lists, so agents keep full working context while critical systems stay provably protected.

Funding

  • OtherFiled Mar 30, 2026Form D
    $35K/ $95K

Product

Kanuki

AI agent security platform

A permission and guardrail platform for AI agents that turns security requirements into explicit no-go boundaries, formally verifies they cannot be bypassed, enforces access within them, and logs all agent actions.

  • Explicit no-go boundary definition for prohibited agent actions
  • Formal verification that boundaries cannot be bypassed, including through indirect paths
  • Visibility into where agents run, what they touch and what access they inherit

Built for

Enterprises deploying AI agents against sensitive internal systems

People on file