Akuity has introduced an Agentic Control Plane and Model Context Protocol server designed to let AI agents participate in software delivery without bypassing the identity, permissions, release checks and audit controls already used by engineering teams.
The launch addresses a practical tension for organizations using AI agents in development and operations. An agent may be able to write code or investigate a deployment, but connecting it directly to a production environment with a token, command-line access or a Kubernetes configuration can create a separate path outside the controls that platform teams use to manage changes.
Akuity’s approach places the agent behind the software-delivery platform rather than giving it a standalone route into infrastructure. The company says the control plane gives agents access to deployment history, cluster health and change lineage, while each request is authenticated as the user who connected the agent. In effect, the agent cannot receive more access than that user has.
The accompanying MCP Server is the interface through which MCP-capable clients can reach the platform. Akuity says it exposes information and tools related to Argo CD, Kargo, fleet insights and its specialized agents. An engineering team could use an agent to examine application health, review changes around an incident, check a release’s eligibility for promotion or prepare delivery configuration for a new service.
For IT teams, the most concrete stated benefit is a governed path for those tasks. Akuity says that actions made through MCP are recorded under the authorizing user or API key and tagged in the audit log. Existing Kargo checks and promotion rules continue to apply whether an action comes from the user interface, command line or an AI agent. Actions requiring approval remain pending until an authorized person approves or rejects them.
The company also says administrators can disable the platform endpoint, disable MCP access for a specific instance, or revoke the user or API credential used by a client. Those controls are relevant for teams that want AI assistance in deployment and incident workflows but need a way to limit scope or remove access quickly.
Akuity positions the product as more than an AI gateway because its platform already holds delivery and infrastructure-operational context. The company says that context can include deployment and promotion events, Kubernetes timeline data, image and CVE information, and audit records. That could reduce the tool switching involved when an SRE or platform engineer is trying to reconstruct what changed before an application became degraded.
In a vendor example, Major League Baseball’s principal DevOps engineer said the control plane identified more than 100 degraded applications shortly after it was enabled and traced a systemic issue to one project in less than 10 minutes. That is a customer account supplied by Akuity, not an independently validated performance measure, and organizations should evaluate the technology against their own environments and approval requirements.
The service is available on the Akuity Platform today, according to the company. Its relevance will be strongest for organizations that want agents to help with production-adjacent work but do not want agent access to become a parallel, less-auditable operational channel. The core decision for buyers is whether Akuity’s existing delivery controls and fleet context match the systems and workflows their teams already use.