Dome Systems launches access to its Agent Operations Platform

REDWOOD CITY, Calif. — September 5, 2026 — Enterprise platform teams can now sign up and start governing their AI agents. Dome Systems, founded by former HashiCorp executives Dave McJannet and Marc Holmes, made self-serve access available today for what it calls the first end-to-end Agent Operations Platform: one system to control the code an agent ships, the tools it calls, and the model it reasons through.

Despite the staggering rate of growth of AI adoption, the deployment of agentic apps is still nascent for most enterprises. And while most conversation in AI tends to be around direct end user productivity — developer harnesses, skills, and the like — the deployment of agents is largely blocked by predictable operations, security and finance constraints.

To date most enterprise efforts have focused on the adoption of point solutions such as gateways and model routers to meet the needs of a particular agent, but the disconnected nature of those offerings prevents platform teams from delivering on consistent capabilities for application development. Policies and controls set at one endpoint don’t carry over to another, and disparate audit systems make observability difficult.

“The landscape is filled with high priced MCP gateways and model routers, each solving one narrow piece of the agent operations requirement,” said Dave McJannet, CEO of Dome Systems. “Tool connections and model brokers are table stakes. Platform teams need a consistent way to govern the entire agent as it iterates through an entire agentic loop, not just the endpoints.”

Dome is focused on governing the agent in its entirety. As an example, a platform team registers a new customer support agent, bringing it under management with credentials provisioned but never held directly. When it reaches the CRM, the gateway decides what the call can do, rules permit or forbid it, guards inspect the response, and quotas cap what it spends. If it needs to reason over sensitive data, the broker routes the call to an approved model provider. With Dome, each decision streams into a single event vocabulary. Across ten teams on different frameworks and clouds, that’s one operational model, with a composable set of controls that can be applied across that agent activity.

Rather than bolting governance features onto a single-purpose gateway or router, Dome built its platform around three connected capabilities:

• A registry that makes every agent known and manageable, with credentials issued, capabilities declared, and lifecycle owned;
• A gateway that evaluates every tool call against policy in full context, allowing, denying, or escalating as needed;
• And a broker that constrains how an agent reasons, through policy-based model selection and routing across providers.

That combination matters most for platform and infrastructure teams overseeing agents across multiple business units, not for a single product team shipping one agent to end users. McJannet has compared the current moment to early cloud adoption.

“Every enterprise building with agents today is assembling a patchwork of gateways and routers, then trying to make them talk to each other,” said McJannet. “We saw this exact pattern with cloud infrastructure, which didn’t truly unlock until the requisite controls were in place to industrialize the application delivery process at scale. Today teams are running three MCP gateways and a model router but still have no consistent way to govern what their agents can do. And offering our platform in a self-service form means any platform team can start fixing that today, without waiting for us.”

With self-serve access live, a platform team can register its first agent under real policy this week rather than waiting on procurement.

Self-serve access to Dome Systems is available now at app.domesystems.ai, with public pricing at domesystems.ai/pricing.