Frontegg.

Here is a fact about software companies that almost nobody says out loud: they keep rebuilding the same login screen. Different logo, same six months. A new SaaS startup gets funded, hires engineers, and one of the first things those engineers do - before the product that justifies the company's existence - is reimplement authentication, password resets, single sign-on, roles, permissions, and the specific hell of multi-tenancy. Then the next startup does it again. And the one after that.

Frontegg's entire premise is that this is insane, and that it should be bought rather than built. The company sells customer identity and access management - CIAM, in the acronym-industrial complex that surrounds security software - as a drop-in layer. You install an SDK, you get a login box, a self-service admin portal, SSO, adaptive multi-factor authentication, role-based and attribute-based access control, SCIM provisioning, audit logs, and the multi-tenant plumbing that makes B2B software work. The idea is that you spend your engineering months on the software only you can make, and rent the front door from someone whose whole job is doors.

The founders came by this belief honestly, which is to say painfully. Sagi Rodin and Aviad Mizrachi met at Check Point, the Israeli cybersecurity giant, in the late 2010s. Rodin was a cloud security director; Mizrachi ran an R&D group. From that vantage they watched team after team inside a large, sophisticated company sink quarters into building authentication from scratch. Rodin asked management to let him assemble a small team to solve it centrally. They built a prototype in a one-day hackathon, then moved from prototype to real implementation - and then Rodin and Mizrachi concluded the problem was bigger than one company.

In September 2019 they left Check Point. They spent roughly a year in stealth: raising about $1 million in pre-seed, making first hires, and assembling a set of pre-built identity capabilities that could slot into any SaaS app. Frontegg officially came out of hiding in 2020, alongside a $5 million seed round led by Pitango. A $25 million Series A followed, and in July 2022 Insight Partners and Stripes led a $40 million Series B - bringing total funding to about $70 million.

What's interesting about the pitch is that it runs directly against a founder's strongest instinct: to build. The urge to write your own auth "because it's core" has torched more roadmaps than any competitor ever has. Frontegg's product is, in a sense, a permission slip to not do the thing - a way to convert a dreaded internal project into a line item. And the market that rewards this is unusually honest about it. Frontegg carries roughly a 4.8-star rating on G2, which is notable given that customer identity is a category users only think about when it breaks. In infrastructure, love is measured in the absence of incidents.

The company also tried to reframe what identity even is. In late 2023 it announced "Frontegg Forward," a repositioning of customer identity from a checkbox - can this person log in, yes or no - into what it called the nerve center of a SaaS product. The argument is that whoever owns the login owns the first and most-repeated moment of the customer relationship: onboarding, self-service, security signals, the works. It's a marketing frame, but it's a defensible one. The login is the one screen every user sees, every time.

Then the ground shifted, as it tends to. AI agents arrived - software that reads your files, calls your APIs, and acts on your behalf, often without anyone quite deciding it was allowed to. This is, if you squint, an identity problem wearing a new hat. An agent that can touch enterprise systems needs to be authenticated, scoped, and logged, the same as any user - except it moves faster, doesn't sleep, and multiplies. Most companies deploying agents in 2025 had, generously, no idea who had authorized what.

Frontegg's response was to extend its identity fabric to the machines. In April 2025 it launched Frontegg.ai, pitched as the first identity management platform built for developers making AI agents. The trick is that it sits on top of the Model Context Protocol, or MCP - the emerging standard, born in the Anthropic ecosystem, that lets agents plug into third-party applications the way a USB port lets you plug in a device. MCP solved connection. Frontegg.ai's claim is to add the part MCP leaves out: authentication, authorization, and a security layer, so every user interacting with an agent gets exactly the right access at exactly the right time.

In November 2025 the company followed with AgentLink, described as an enterprise-grade MCP server that lets a SaaS application safely expose its APIs to AI interfaces such as ChatGPT, Claude, and Gemini. If MCP is the door AI agents walk through, AgentLink is the bouncer - checking identity before an agent touches a single endpoint. It is, characteristically, the boring and essential half of an exciting idea.

Whether "identity for humans" and "identity for agents" are one market or two is the open question Frontegg is now betting on. The company is small - around 74 people - competing in a crowded field against Auth0 (owned by Okta), WorkOS, Okta itself, FusionAuth, Descope, and Stytch. But its wager is coherent: that the same unglamorous discipline that made login boring and reliable for people is exactly what the chaotic new world of AI agents is going to need. And that, once again, nobody should have to build it twice.