It used to be that when you started a new job, someone handed you a laptop, a badge, and logins to the handful of systems you needed. You probably didn't get admin rights to payroll. And if you accidentally deleted a shared folder, IT could see exactly who did it.

AI agents start work at companies every day now, and most of them skip that whole process. They show up with whatever access the person who set them up happened to have, and when something goes sideways, there's often no record of which agent did what. (Ask a security team how they feel about that… then give them a minute.)

Mintive handles onboarding for agents. Today the company announced it now handles 2.5 million agent calls a day across more than 50 paying customers. It also announced its $11M funding led by Cowboy Ventures. Mintive's investors include Coatue, Operator Collective, Burst Capital, Maven Ventures, and Hustle Fund, along with Okta co-founder Frederic Kerrest, WorkOS founder and CEO Michael Grinich, Discovery Loop co-founder and former Google Chief Scientist Jeff Dean, and OpenAI co-founder Andrej Karpathy. We're proud to be in that group.

‍

What does Mintive do?

Mintive is the infrastructure for the agentic enterprise: with five products (MCP Gateway, LLM Gateway, Agent Identity, Agent Monitor, and Coworker agents) working together to accelerate enterprise AI adoption. 

Mintive gives every AI agent its own identity, its own permissions, and an audit trail, the same way companies already treat employees. Companies route their agents through Mintive, and security teams get one place to set the rules, whether an agent runs on Claude, ChatGPT, Copilot, Cursor, or something built in-house.

The team started by helping enterprises secure MCP (Model Context Protocol, the standard way AI agents plug into tools like Slack, GitHub, or a company database). Customers kept asking for more, and Mintive now covers the whole job: who each agent is, what it can touch, and what it does.

Co-founders Jiquan "JQ" Ngiam (CEO) and Vijay Vasudevan started the company after watching the first enterprise agents go into production with broad access and no record of their actions. Customers include Ro, Coursera, Modern Treasury, the University of Michigan, and Apax.

‍

Why should you pay attention?

Companies are adding agents fast. Last year, Gartner predicted that 40% of enterprise applications would include task-specific AI agents by the end of 2026, up from less than 5% in 2025. Earlier this year it went further, projecting that the average global Fortune 500 company will run more than 150,000 agents by 2028, up from fewer than 15 in 2025. Only 13% of organizations believe they have the right governance in place for them, and Gartner's own advice for handling the sprawl includes giving every agent a defined identity and set of permissions.

That's a lot of new coworkers who never sat through security training.

Every agent platform comes with its own permissions model, its own integrations, and its own controls. A company running coding agents in Cursor, a chat assistant in ChatGPT, and a homegrown agent for finance ends up with three separate rulebooks. Security teams are left with two bad options: block the tools, or let agents run with wide access and hope for the best.

Blocking rarely works anyway. As Christian Burrows, Head of Security at Stability AI, puts it: "When employees want to use AI agents, simply blocking access isn't an effective security strategy. People will find ways around it."

‍

How does Mintive actually work?

Mintive sits between a company's agents and everything those agents touch. 

• MCP Gateway controls which tools an agent can reach, like Salesforce, a code repository, or an internal database, and what it can do once it gets there
• LLM Gateway controls calls to AI models, so the same rules apply whether an agent runs on Claude, GPT, or something else
• Agent Identity gives each AI agent its own scoped, revocable credential, with access delegated from the person it works for.
• Agent Monitor records the chain behind every action: the person who initiated it, the agent that acted, the server it reached and the model involved. Teams can track spend by user, model and tool, and review any session call by call with timestamps, arguments and results.

Every tool call traces back to four things: the person who kicked it off, the agent that acted, the system it reached, and the model involved. When someone asks who changed a record, the answer is on file.

Because the controls live in one layer, a company can adopt the next agent platform without rebuilding its security setup. The platform is SOC 2 Type II audited.

Mintive also runs agents of its own. Its Coworker agents work alongside teams in Slack or Teams, writing and reviewing code, helping sales and marketing, and pulling from company data. They have long-term memory, so they pick up where they left off and take on work without waiting to be asked.

Customers are already using Mintive in settings where mistakes are expensive:

• Modern Treasury built a bank operations agent with Mintive that searches Sentry, Notion, Linear, the codebase, and internal tools before a technical account manager steps in. It cuts 30 minutes from simpler cases and up to four hours from complex ones.
• The University of Michigan uses Mintive as its MCP gateway, giving 150,000 students, faculty, and staff one route to approved tools while central IT manages access in one place.
• Ro runs Mintive in a healthcare environment, where every agent action has to meet security and compliance requirements.

‍

Why is OpCo invested?

JQ and Vijay have built AI infrastructure at serious scale. JQ was part of Coursera's founding team and later a senior staff software engineer at Google Brain. Vijay co-created TensorFlow, Google's open-source machine learning framework. They know what it takes to build software that large organizations trust with their real systems.

Mintive has handled more than 200 million agent calls in seven months and now runs 2.5 million a day. Its 50+ paying customers span healthcare, payments, education, and private equity - all industries where security reviews are long and strict.

Mintive also gets stickier over time. Once it's in place, it sits in the path of every tool call, and each new connector, policy, and month of audit history makes it harder to replace. It works across every model, so customers can switch or add models without switching vendors.

Identity for people grew into one of the biggest categories in enterprise software. Agents are on track to outnumber the people. 

‍

What's next for Mintive?

If your team is putting agents to work and wants one place to manage who they are and what they can access, visit mintive.com to learn more, or reach out to our team for an introduction.