← All companies
Generate ideas →
MIT
MIT
Agentic Fabriq
ActiveOkta for Agents.
Winter 2026Founded 20252 peopleSan Francisco, CA, USA
AI insightcan contain mistakes
Agent Data PermissioningSaaSEnterprise IT and security teamsMedium competition
Moat
Unified permissioning model for humans and AI is novel; integration breadth creates stickiness.
Key risk
Identity vendors (Okta, OneLogin) may integrate agent permissioning; adoption depends on agent deployment.
Why now
Enterprise AI agent adoption is surging; data governance for agents is critical.
Competitors
Okta, OneLogin, custom IAM
About
Agentic Fabriq is the single control plane for data permissioning throughout your entire organization across your AI and employees. Through Fabriq, any employee can tap into AI/agents with exactly the data and tools they're cleared to use by an admin. Fabriq can be used internally with employees or externally for customers while handling the auth + integrations + permissions securely for you.
Founders · 2
Paulina XuFounder
AI + Physics @ MIT, working on Agentic Fabriq (security middleware for all agentic communication)
Matthew XuFounder
Dropped out of MIT at 19. Building Agentic Fabriq, the security layer between agentic applications and tools.
Related startups
MultifactorFall 2025
Zero-trust authentication, authorization, and auditing for AI agents
Active
Agentin AIWinter 2025
AI Agents that automate enterprise software processes
Active
