The open architecture for a multi-model world
Stacklok is the alternative to vertically-integrated agent stacks from frontier labs and hyperscalers.
Frontier labs have been setting the pace of enterprise AI adoption, but the ground is shifting.
Open weight models have closed the performance gap. Enterprises are pivoting from tokenmaxxing to tokenomics. And agents are evolving from simple and scoped to dynamic and chartered with complex actions.
We’re at the point where enterprises must avoid frontier lab lock-in and embrace the optionality of an open architecture.
Leaders across industries trust Stacklok’s open architecture
Control your Costs
“Everything at project44 is Kubernetes-native. Since all of our services run on Kubernetes, it’s efficient to run our MCP servers the same way.”
Control your data
“Our goal is to make everyone at TRM a builder. Stacklok’s agent runtime ensures the underlying infrastructure is secure and reliable at scale.”
Control your destiny
“Our partnership with Stacklok is about building a platform that meets the realities of enterprise AI today, and the emerging future.”
Here’s why we think you should consider Stacklok
We co-created Kubernetes to help enterprises embrace cloud native infrastructure. At Stacklok, we’re helping enterprises use that infrastructure to run and govern the next generation of agents.
Our motivation is simple. The AI landscape is in constant motion. Your agents are evolving from tightly scoped to highly dynamic, and the right model for a given task is always changing. In this environment, enterprises need optionality.
Meanwhile, the frontier labs and hyperscalers are all building vertically-integrated, closed stacks. The experience is so good, you might not even notice the walls … until you find yourself locked in.
So we’re building an open architecture that is:
Open to enable you to mix models, providers and components, with an open source core that you can make your own.
Self-hosted to ensure your data always stays behind your firewall.
Kubernetes-native to deploy rapidly, integrate with your existing IdP and manage agents the same way you manage your other workloads.
Identity-first to connect every agent action, tool call and model call to the responsible individual or team.
Our customers are using their existing cloud native infrastructure to support their next generation agents, while staying in full control of their costs and their destiny.
We’re only just getting started. The community is vibrant and growing and we invite you to dive in.
By the numbers:
Stacklok secures 10M+
tool calls every month (and growing)
We’ve built the gateway bookends to secure every agent-to-tool and agent-to-model action and the first cloud-native agent harness, so you can run at scale.
MCP Gateway
Ensure every agent-to-tool action is approved and secured
Deploy an outbound-only tunnel to keep your data and systems behind your firewall
Agent Harness
Explore the first cloud-native harness that supports fleets of agents
Assemble your harness from our library of tools and skills to get up-and-running fast
LLM Gateway
Protect every agent-to-model action and regain control of AI costs
Don’t delegate model selection criteria; define policies that meet your needs
Common identity, audit, observability and policy
Just the facts
Everything you wanted to know about Stacklok:
Craig McLuckie (CEO) and Joe Beda (CTO) co-created Kubernetes (and Craig founded the Cloud Native Computing Foundation)
We maintain an open source MCP platform, ToolHive, and cloud-native harness, Mecatl
We’re charter members of the Agentic AI Foundation
Stacklok is a maintainer of the official MCP registry
We have an amazing team that spans 10 countries
10M+
Our platform secures
10M+ tool calls every month
Stacklok is well funded by Accel, Bain Capital Ventures and Madrona
Ready to see it live?