Today we are launching Mintive, the infrastructure for the agentic enterprise. Mintive provides an MCP gateway to govern tools, an LLM gateway to control models, and infrastructure to deploy agents. We treat each agent the way an enterprise treats an employee: it gets its own identity, only the access its work requires, and a record of everything it does, regardless of the model or tool underneath.

We're also announcing that we've raised $11 million in total funding. Our latest $7 million round was led by Cowboy Ventures, with new investors Operator Collective and Burst Capital joining and existing investors Coatue, Hustle Fund and Maven Ventures participating. Our angels include Frederic Kerrest, co-founder of Okta; Michael Grinich, founder and CEO of WorkOS; Scott Belsky, former Chief Product Officer at Adobe; Jeff Dean, co-founder of Discovery Loop and former Chief Scientist at Google; and Andrej Karpathy, co-founder of OpenAI.

Mintive is in production with more than 50 paying customers, including Ro, Modern Treasury, Braze, Coursera, the University of Michigan, Apax and Stability AI. We process over 2.5 million agent calls each day, and have handled more than 200 million in seven months.

What enterprise AI is missing today

Eighteen months ago, AI was an advisor: you asked, it answered, you did the work. It soon became a personal coworker, connected to your tools and taking tasks off your plate. This year, agents became teammates, with memory and the ability to act on their own.

Gartner expects the average Global Fortune 500 company to run more than 150,000 AI agents by 2028, up from fewer than 15 in 2025.

We see the agentic enterprise taking shape in four levels. At each one, agents act with more autonomy, and the infrastructure underneath has to grow with them.

AI writes better docs and emails for you. You ask, it answers, you do the work.

The challenge

Someone pasting a separation agreement into a consumer chatbot forces the company to decide which models people can use and what they can put in them.

Infrastructure it needs

1. LLM GatewayApproved models and one policy across providers

AI completes real tasks for one person at the company, connected to real systems.

The challenge

Every company has a folder of sensitive information no model should see. But each connector signs in its own way, and agents pick up the full access of whoever set it up.

Infrastructure it needs

1. MCP GatewayEvery connector behind one door, scoped by role
2. LLM GatewayApproved models and one policy across providers

Shared agents show up in Slack and Teams, with memory and the ability to act.

The challenge

Coworker agents live in your Slack and Teams channels, where many people and many agents can share one thread. What an agent may do depends on the channel, the conversation, the task and who is messaging it. A contractor without an NDA asking for the customer list should get a different answer than the CFO asking in the same thread.

Infrastructure it needs

1. Agent MonitorThe chain from person to agent to action
2. Agent IdentityScoped, revocable credentials delegated from a person
3. Coworker agentsPermitted into rooms explicitly, answering within the asker’s access
4. MCP GatewayEvery connector behind one door, scoped by role
5. LLM GatewayApproved models and one policy across providers

Agents run whole workflows with minimal human input and hand work to other agents.

The challenge

Agents handing work to other agents turn one request into a network of steps no one can follow end to end. Orchestrating that network, and understanding how it got the work done, becomes the hard part.

Infrastructure it needs

1. OrchestrationOne view of how agents hand off work and get it done
2. Agent MonitorThe chain from person to agent to action
3. Agent IdentityScoped, revocable credentials delegated from a person
4. Coworker agentsPermitted into rooms explicitly, answering within the asker’s access
5. MCP GatewayEvery connector behind one door, scoped by role
6. LLM GatewayApproved models and one policy across providers

We started Mintive after seeing the first enterprise agents enter production with broad access and no clear record of which agent had taken which actions. As companies add agents across teams and departments, each new agent widens access the company cannot see or control.

"We're early in the enterprise adoption of AI with the major gating factors being governance, IT and security team ability to oversee, identity management, and security. Mintive offers the critical, scalable, trusted third party scaffolding businesses need - to deliver agent identity, permissions and tracing, regardless of the underlying models."

Aileen Lee
Founder and Managing Partner
Cowboy Ventures

Five products for deploying and managing agents

Mintive has five products that work together:

• MCP Gateway governs tool access: approved servers, role-based bundles, tool-level switches. Every connector sits behind one door, and only the enterprise's AI clients get through it.
• LLM Gateway governs model calls: it routes traffic through approved models, sets budgets by group, and applies one policy across providers.
• 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.
• Coworker agents work alongside your teams in Slack and Teams, with long-term memory and the ability to get work done proactively. They are permitted into rooms explicitly.

Coworker agents sit in this stack to accelerate AI adoption: they run on the same identities, permissions and audit trail as the gateways, so end users can put agents to work the day IT and security teams say yes.

"When employees want to use AI agents, simply blocking access isn't an effective security strategy. People will find ways around it. The better approach is to make saying yes the easy path, giving employees the freedom to use these tools while ensuring security teams have the visibility and control they need over what agents can access and do and this is what Mintive does for us."

Christian Burrows
Head of Security
Stability AI

The controls that decide what an agent can reach and do sit with the company, outside any model. Mintive is model-agnostic: adopt Claude, ChatGPT, Copilot, Cursor or your own agents, and the identities, permissions and audit trail stay the same underneath. Mint Guard screens calls for prompt injection, exposed secrets and personal data before they reach company systems. Every call is logged, and the platform is SOC 2 Type II audited, with EU and US data residency.

Our customers

At Modern Treasury, a bank operations agent searches Sentry, Notion, Linear, the codebase and internal tools before a technical account manager steps in. The deployment went live in under three weeks and saves their technical account managers 30 minutes to 4 hours per case.

At Ro, Mintive operates in a healthcare environment where agent access and auditability carry security and compliance requirements alongside the work itself.

At the University of Michigan, Mintive is the enterprise MCP gateway across 150,000 community members. Students, faculty and staff have one path to approved tools, and central IT controls access in one place.

"Mintive brings together information from authorized services such as Google Workspace, Canvas, Slack, Zoom, and Dropbox, subject to users' existing access permissions. This reduces the need to search across separate applications and support teaching, learning, research, and collaboration in a large, complex university environment."

Robert Jones
Assistant Vice President of Emerging Technology and Support Services, ITS
University of Michigan

We've helped teams like Apax Partners, Ro and Stability AI roll out company-wide in days.

If you're a leader in AI, security or IT facing the same challenges, book a demo.

Build with us

We are a team of scientists and builders from Stanford, Google Brain, Coursera, Applied Intuition, Snorkel and Amazon, working alongside our own agents. We are hiring across Engineering, Marketing and GTM. If you want to work on these problems, see our careers page or email hiring@mintive.com.