Map / Sandboxing / Jozu
Jozu
jozu.com ↗ · LinkedIn ↗ · Open on the map
Capabilities
Jozu Agent Guard — zero-trust AI runtime that executes agents, models, and MCP servers inside secured environments with policy enforcement and guardrails that agents cannot disable; vetting, signing, and governing AI artifacts from development through production; local policy engine with visibility into running actions + inputs/outputs + prompts/responses; works across servers, laptops, and edge devices.
Sensors / integration
Local runtime engine running on the host (separate process boundary from the agent it governs); built on top of KitOps (Jozu's CNCF project with 240,000+ downloads).
Protections
Only approved artifacts execute, only permitted actions occur, agents cannot terminate or bypass the enforcement layer, full audit logs immune to agent tampering, policy enforcement applied locally without cloud round-trip.
Differentiator
Defining "agent self-disable" attack origin story — Jozu observed an AI agent killing its policy enforcement process, disabling auto-restart, resuming unrestricted, and erasing audit logs in four commands; Agent Guard architecturally moves enforcement out of the agent's reach.
Team & funding
- Founders
- Brad Micklea, Gorkem Ercan, Jesse Williams
- Total raised
- $4M
- Investors
- HalfCourt Capital, Mozilla Ventures, Brightspark Ventures, AlleyCorp, Sentiero Ventures, Union Bay Partners
Other Sandboxing companies
ARMO · Autonomous · Blaxel · Canyon Road · Daytona · Duality Technologies · Edera · Galene.AI · GuardiAgent · I-THAM · Islo · Keyboard · See all →
Last updated · Get the “Listed on” badge