No items found.
No items found.

Introducing Solo Enterprise for agentgateway on Any Infrastructure

Solo Enterprise for agentgateway now runs on Kubernetes or any infrastructure, with standalone deployment, full UI management, and enterprise MCP capabilities.

Agentgateway has seen tremendous growth this year with more than 50 million image downloads, over 500 contributors, and the project’s addition to the Agentic AI Foundation. Adoption at this scale presents new and exciting opportunities for agentgateway to help customers with a diverse range of deployment environments and operating models.

Solo Enterprise for agentgateway has always run on Kubernetes, and for good reason. Kubernetes gives the control plane and data plane high availability, horizontal scale, and a GitOps operational model where configuration is code and every change is reviewed and versioned. 

For a platform team, this is the right foundation, and it isn't changing. But many organizations take their first steps on VMs, ECS, or even bare metal. In addition, the personas actively configuring and managing agentgateway require the full set of capabilities to be available through a self-service UI, rather than GitOps and configuration as code. 

Solo’s vision for agentic infrastructure extends to all environments and personas. Today, we are excited to announce that Solo Enterprise for agentgateway now comes in two modes.

Kubernetes mode is the deployment you already know. It's a control plane built for Kubernetes, designed for teams with a platform team and a GitOps workflow.

Standalone mode is new. It's a self-contained enterprise distribution that runs anywhere, with a full read-write UI for configuring and managing the gateway.

Same product. Same policy model. Same identity model. Same MCP gateway capabilities. You just choose the mode that fits how your team actually operates.

Run it anywhere 

Imagine a company where every employee just got a Claude Enterprise seat. IT owns the rollout and the security governance requirement. Getting agentgateway in front of this traffic is the obvious solution, but it needs to fit in the organization’s existing IT infrastructure and procedures which don’t include Kubernetes today.

That's who standalone mode is for. And there are plenty of other teams in the same situation:

  • Small teams moving fast. A few engineers building agentic applications who want a gateway next to the rest of their stack. No cluster. No Helm charts.
  • Centralized gateway deployments. One large, well-provisioned host sitting in front of the business, managed through a UI. That's a completely reasonable way to run a gateway.
  • Edge and embedded environments. Ships, factory floors, retail back offices, remote sites. Places where there may not be a platform team nearby, or even a Kubernetes cluster.

In summary, If you can run a process, you can run agentgateway.

Configure everything through a UI

Standalone mode comes with a full management UI.

Configure your model providers. Connect your MCP servers. Define policy. Inspect traces. See cost and model attribution for individual requests, and you can do all of that without touching a YAML file. 

Additionally, configuration lives in a shared, database-backed store. That gives you self-service through the UI without giving up high availability and resiliency.

What's under the hood

Agentgateway open source has shipped standalone mode from the start. Solo Enterprise for agentgateway builds on that mature foundation.

On top of it, the enterprise distribution adds:

  • MCP composition. Combine several tools into one. Agentgateway chains the steps for you, so you skip building a custom server to connect them.
  • Code mode. The agent writes a short script to use your tools, instead of calling them one at a time. This keeps it fast even when you have many tools.
  • Search mode. The agent looks up the tool it needs, instead of loading every tool at once. This also keeps things fast with a large tool catalog.
  • Brokered elicitation. The gateway asks the user to log in, then holds that login for later use. The MCP server never sees the user's credentials directly.
  • Impersonation. The gateway swaps the user's login for its own trusted credential. That new credential still stands for the same user.
  • Delegation. The gateway issues a credential that names both the user and the agent working on their behalf. Anyone who checks the request sees both identities.
  • Dual authentication. The gateway checks the agent's identity first, then checks it again with a trusted credential before the request reaches your systems.

See the full feature list here.

Kubernetes mode or standalone mode?

There is no need to choose.

Adopting Solo Enterprise for agentgateway means you can run Kubernetes mode, standalone mode, or both. Pick based on what your organization needs.

You get the same capabilities either way. The form factor flexes to meet every environment and every team where they are.

Get started

Solo Enterprise for agentgateway is available now in both modes.

Grab the standalone distribution, point it at your model providers and MCP servers, and put an enforcement point in front of your AI traffic in minutes.

Get started by visiting our docs.