Agentic Fabriq

Problem/Solution Report

Problem Enterprises are rapidly deploying AI agents, but existing identity and permission systems are built for humans, not autonomous agents. This creates over‑permissioned agents, data‑leakage risks, and fragmented integrations that are hard to audit. Companies report that a large majority have experienced "agent incidents" where agents hallucinate, invoke unauthorized APIs, or mishandle sensitive information.

Solution Agentic Fabriq delivers a secure middle‑layer – a single control plane that enforces identity, per‑user permissions (acts‑as‑user), least‑privilege defaults, tenant guardrails, SSO/IdP integration, pre‑built connectors, and full audit logs for every agent action. This platform centralizes policy, lets IT register tools and agents, and ensures agents cannot exceed the privileges of the invoking employee.

Value Proposition

• Speed to integration: Connect tools in minutes without custom auth code.
• Security: Per‑employee, least‑privilege access prevents agents from over‑reaching.
• Compliance & Visibility: Centralized audit logs provide traceability for regulators and internal governance.
• Positioning: Marketed as "Okta for agents," the solution mirrors proven identity‑management models for the emerging agentic AI stack.

Go‑to‑Market Backed by Y Combinator (Winter 2026) and operating out of Sunnyvale, the company is in an early‑stage, enterprise‑focused sales model with pricing available on request, typical of B2B SaaS pre‑product‑market fit.

Market and Competitors Report

Market Context The addressable market sits at the intersection of enterprise Identity & Access Management (IAM/PAM) and the fast‑growing Agentic AI stack. IAM is a steady‑growth market (~USD 22 B in 2025 → USD 78 B by 2034). Agentic AI is expanding rapidly (CAGR > 40 %), projected to reach USD 139 B by 2034. Enterprises will need per‑user, least‑privilege controls for agents that act across SaaS tools and internal systems.

Direct / Near‑Direct Competitors

• Credal.ai – AI security layer providing access controls, data masking, and audit logs.
• WitnessAI – Governance and compliance platform for enterprise AI usage.
• Keycard – Deterministic identity and access stack for AI agents with per‑task permission boxes.
• Other LLM‑security vendors (Lakera, CalypsoAI, WhyLabs/Lasso) are adding authorization and audit features but focus mainly on prompt‑injection and data‑leakage.

Adjacent Authorization Infrastructure

• WorkOS FGA, Oso, Cerbos, Open Policy Agent (OPA), OpenFGA – Fine‑grained policy engines that require custom integration and do not natively offer an "acts‑as‑user" model for agents.
• Traditional IdPs (Okta, Azure Entra ID) and cloud AI platforms (OpenAI Workflows, Azure/Google AI) may extend into agent‑specific RBAC but currently lack a unified cross‑tool control plane.

Agentic Fabriq Positioning Fabriq differentiates by providing a managed, security‑first control plane that:

• Enforces per‑employee permissions for agents (acts‑as‑user).
• Offers out‑of‑the‑box connectors and tenant guardrails.
• Delivers full auditability across all agent actions.
• Integrates with existing IdPs rather than replacing them. This reduces the engineering burden of building custom authz for each agent/tool combination.

Risks & Considerations

• The space is nascent; larger IdPs and cloud AI providers could add native agent permissioning.
• Success depends on breadth of connectors, performance at machine‑speed, and clear policy authoring UX.
• Competitive pressure from both dedicated agent‑authz startups and broader governance platforms.

Founders and Background Report

Paulina Xu – Co‑founder & CEO

• Studied AI and Physics at MIT, conducting research at the MIT Kavli Institute, MIT Haystack Observatory, and Italy’s INAF Padua.
• Her work focused on applied AI, scientific computing, and rigorous data workflows, giving her deep expertise in building secure, auditable systems.
• Dropped out of MIT in her second year to launch Agentic Fabriq and participated in Y Combinator’s Winter 2026 batch.
• Listed on the company’s About page and on Crunchbase as Co‑founder/CEO, with Y Combinator as lead investor.

Matthew Xu – Co‑founder & CTO

• Studied AI and Mathematics at MIT, serving as a research assistant at MIT CSAIL and the MIT FutureTech Lab.
• Built RNA‑seq analysis pipelines and analyzed large‑tech AI investment data, blending software engineering with data‑infrastructure expertise.
• Also left MIT early to co‑found the company, joining the same YC Winter 2026 cohort.
• Appears on Crunchbase and the YC profile as Co‑founder/CTO.

Shared Background

• Both founders met at MIT, share a technical foundation in AI, and have experience turning research prototypes into production‑grade systems.
• Their combined expertise in AI research, data pipelines, and systems engineering underpins Agentic Fabriq’s focus on “Okta for agents” – a secure control plane for AI agents.
• The company is headquartered in Sunnyvale, CA, and is backed by Y Combinator.