Permit.io

Researched 2026-06-28. Primary category: authorization-engine. Independent, VC-backed (Series A).

One-liner — Developer-first authorization-as-a-service: a hosted policy decision point plus SDKs and a no-code policy editor, so engineers add fine-grained access control (RBAC/ReBAC/ABAC) without building it from scratch — increasingly aimed at AI agents and MCP.

What it does

Permit.io wraps proven open-source engines (open-policy-agent / OPA and AWS Cedar) behind a managed control plane and SDKs. You model roles, resources, and relationships in a UI or as code; Permit distributes policy to PDPs you run or it hosts, and your app/agent calls a simple permit.check() for allow/deny. It also maintains OPAL (Open Policy Administration Layer), its open-source project for real-time policy/data sync to OPA. Newer products (Approval Flows, agent authorization) target AI bots and agents asking for fine-grained, identity-aware tool access.

Where it sits in the stack

Primary authorization-engine; cross-listed to mcp-gateway because it pitches agent/MCP tool-access authorization. Layer: model-prompt. Lethal-trifecta role: constrains sensitive-data and egress by deciding whether an agent’s action is permitted per request; not prompt-content inspection (that’s ai-runtime-security / agent-runtime-security). Building block of trust-zone-segmentation; yellow/green zones.

Deployment & architecture

Hybrid: Permit hosts the control plane (policy authoring, audit) while the PDP can run locally (sidecar/container) so authorization decisions and sensitive data stay in your environment — useful for regulated shops. SDKs across major languages; REST API for policy management. Built on OPA/Cedar + OPAL. Integrates with existing auth/IdP (it authorizes, it doesn’t authenticate).

Positioning & differentiators

Known for developer experience and speed-to-implement — no-code editor plus code, and “don’t write your own authz.” Versus open-policy-agent (raw engine, write your own Rego), styra (enterprise OPA management), cerbos (stateless OSS engine, also hosted), oso (Polar DSL), authzed (Zanzibar/ReBAC graph). Permit’s hedge is engine-agnostic (OPA and Cedar) with a managed wrapper; its agent/MCP messaging is early but active.

Ownership, funding & M&A

Founded 2021 (originally “Authorizon”) by Or Weis (CEO) and Asaf Cohen; HQ Tel Aviv. Raised ~$14M: ~$6M early round then Series A $8M (2024-02, led by Scale Venture Partners, with NFX, Firestreak, Roosh, Verissimo). No M&A — independent (no seed flag). Confidence high.

CTO / hedge-fund lens

Day-2. Good fit when you want externalized, fine-grained agent/tool authz without standing up and operating OPA+Styra yourself — the local-PDP model keeps decisions and data in-house, which matters for a fund. Lower engineering lift than raw OPA; more opinionated than Cerbos. Diligence the maturity of the agent/MCP features (newer than the core authz product) before betting an agent program on them. Series-A vendor — weigh vendor-risk for a core control.

Competitors / alternatives

open-policy-agent, styra, cerbos, oso, authzed, pomerium.

Open questions / to verify

• Maturity/production references for the agent / MCP authorization features specifically (vs core app authz).
• Any funding since the 2024 Series A.

Sources

• Permit.io Raises $8 Million for Authorization Platform (SecurityWeek) — fetched 2026-06-28 — supports: Series A $8M 2024-02 Scale VP, ~$14M total, Tel Aviv, founders, full-stack authz, agent/AI support; confidence: high

History

• [2026-06-28] Stub created from seed registry.
• [2026-06-28] Researched; established 2021 founding (ex-Authorizon), Tel Aviv HQ, founders (Weis/Cohen), ~$14M funding (Series A $8M 2024 Scale VP), OPA/Cedar + OPAL architecture, agent/MCP positioning. No M&A — independent. Confidence high.