Mintive Raises $7 Million to Give AI Agents Their Own Identities
Mintive, an agent identity startup, has announced a $7 million seed round led by Cowboy Ventures, bringing its total funding to $11 million. TechCrunch reported the round on October 29, 2026.
Coatue, Operator Collective and Hustle Fund also invested in the company. Individual angel investors include Frederic Kerrest, co-founder of Okta, former Google chief scientist Jeff Dean and Andrej Karpathy.
Mintive was founded by Jiquan Ngiam, who is CEO, and Vijay Vasudevan, who is CTO. Ngiam was the founding engineer and first employee at Coursera, hired by Andrew Ng, and he worked on neural networks at Google Brain. Vasudevan co-created TensorFlow, the machine learning framework, while at Google Brain.
The platform combines several components. An MCP gateway (MCP, or Model Context Protocol, is an open standard that lets agents reach outside tools and data) controls which cloud services an agent may access and which actions it may perform. An LLM gateway routes requests to different large language models and can cap spending. An agent monitor tracks every agent across a company. The platform also manages agent identities and authorizations, which can differ from those of the employees the agents serve. Separately, Mintive's website describes coworker agents that act as AI teammates in Slack and Teams.
Mintive launched its product in February. The company says it has 50 enterprise customers, including Ro, Modern Treasury, the University of Michigan, Coursera, Apax and Stability AI.
The broader context here is the mismatch between how enterprises have modeled identity and how agents now operate. Directory services, single sign-on and role-based access control were designed around people who log in, hold roles and eventually leave. An agent that calls APIs, reads internal documents or posts to a shared channel may run continuously, act for several users and hold permissions that no single employee record describes. Treating that agent as a service account, or as a copy of its human owner, creates audit and blast-radius problems (the damage a compromised credential can do) that conventional tooling was not built to answer. Mintive is aiming at that gap with an identity and authorization layer for agents, paired with the gateways and monitoring needed to see what they do and what they cost.
In my view, the useful question for buyers is less whether an agent has an identity than whether that identity's scope can be expressed, reviewed and revoked as quickly as the agent's behavior changes. A policy layer is only as sound as the rules written into it, and a cost cap or access list still depends on the organization maintaining it. The customer list spans fintech, higher education, private equity and generative media, which suggests the problem is not confined to one vertical.
If the product performs as described, the practical benefit is visibility. Security teams could see which agents exist, what they touch and what they spend, rather than discovering them after the fact. Developers could use a single gateway for model routing and tool access, reducing the number of bespoke integrations each team builds. Those gains would matter most to organizations already running agents in production that have not yet decided how those agents should be governed.
