Mintive raised a $7 million seed round to grow its platform, which gives each AI agent inside a company its own identity, scoped permissions and an audit trail.
Mintive, the startup that gives every AI agent inside a company its own identity, has raised a $7 million seed round led by Cowboy Ventures. Coatue, Operator Collective, Burst Capital, Maven Ventures and Hustle Fund also took part. This brings Mintive’s total funding to $11 million.
The angel list is the eye catching part. Okta co-founder Frederic Kerrest, WorkOS CEO Michael Grinich, Scott Belsky, former Google chief scientist Jeff Dean and OpenAI co-founder Andrej Karpathy all wrote checks.
Until recently the company went by MintMCP. Fair enough, since that’s where it started.

MCP (Model Context Protocol) is the open standard that lets AI models plug into outside tools and data, like a CRM, a code repo or a shared drive. It’s handy. It’s also a bit of a mess for security teams. When an employee sets up an agent, the agent often just inherits that person’s access, which tends to be a lot broader than the agent actually needs. Spread a few hundred of those across Slack, coding tools and chat apps and nobody can say for sure which agent did what, or who signed off on it.
CEO Jiquan “JQ” Ngiam’s pitch is pretty simple. Companies give each employee a login and a set of permissions, he said, and each agent needs the same treatment.
So Mintive treats every agent as its own user. Each one gets scoped credentials and short lived tokens that can be pulled at any time, and everything it does lands in an audit log. The MCP gateway decides which tools and cloud services an agent can reach. The LLM gateway sets which models a team can call and caps what it spends on them. An agent monitor shows activity across the whole company. Then there’s Mint Guard, which screens tool calls for prompt injection, exposed secrets, personal data and harmful content (some of that gets blocked outright, some only gets logged). Mintive also sells its own coworker agents that sit inside Slack and Teams.
The product launched in February. Since then Mintive has signed more than 50 paying enterprise customers, including Ro, Modern Treasury, Coursera, Stability AI, Apax and the University of Michigan. Together they run about 2.5 million agent calls a day, and more than 200 million over seven months. Michigan uses it as a shared MCP gateway run by central IT. Modern Treasury says it saves around 30 minutes on simpler cases and as much as four hours on the hard ones.

Recommended: Pennant Raises $5M In Seed Funding Round
Customers tend to be a bit surprised at how fast it is to set up, Ngiam said.
The founding team has serious machine learning credentials. Ngiam was the founding engineer at Coursera (now a customer) and later a senior staff software engineer at Google Brain. CTO Vijay Vasudevan helped create TensorFlow there. Bozhi See and Phillip Bensaid round out the co-founders.
Cowboy Ventures founder Aileen Lee said the industry needs a lot more infrastructure that gives companies governance, visibility and control over what their agents are doing.
The plan with the funding is to keep building that out: one identity per agent, tight permissions and a record of every action, across whatever tools and models a company runs.
It won’t be quiet out there. Amazon, Microsoft, OpenAI and Anthropic all have their own agent management tools. LangChain, New Relic and Rippling sell narrower products that cover pieces of the problem. And more than 40 Y Combinator startups are chasing some version of the same idea.
Please email us your feedback and news tips at hello(at)techcompanynews.com

