Akuity launches control plane for AI software delivery

Akuity has launched an Agentic Control Plane and MCP Server for software delivery, designed to govern how AI agents operate across delivery pipelines.

The launch addresses a growing issue for engineering teams as AI tools move beyond writing code into managing production changes and release processes. Many agents already operate in production environments with limited access to operational data, while some connect through open-source MCP servers that provide direct access without inherited identity, policy controls or audit trails.

The new products are intended to sit inside the software delivery workflow rather than alongside it. Akuity says its control plane runs the delivery pipeline itself, giving it access to deployment history, cluster health and change lineage across environments.

That operational view is central to the company's pitch. Rather than limiting agents to application-level information, the platform is meant to let them examine infrastructure state, review recent changes and act within the same controls already applied to human users.

Operational controls

Every request from an AI agent is routed through the Agentic Control Plane and authenticated as the specific user who connected the agent. Akuity says this means the agent inherits the user's identity and permissions across the platform rather than relying on a separate set of credentials.

The policy layer can also limit what an agent is allowed to do, including restricting sensitive production actions even if the underlying user would otherwise have access. Every action is recorded through the same audit trail used elsewhere in the delivery process.

The MCP Server is designed to provide access for in-house and third-party agents, including Claude, Codex and Cursor. Through the server, those agents can query cluster state, inspect deployments and trace software changes back to the underlying code commit.

Use cases

Akuity outlined several tasks it expects customers to hand to agents through the new system. These include checking the condition of multiple clusters through a single query, setting up delivery workflows for a new service from an empty repository, and reviewing whether a release qualifies for promotion to the next environment.

Users can also ask an agent to investigate a degraded system. In that scenario, the user's agent can connect with Akuity's On-Call Agent, gather logs and recent change data, and recommend a fix.

The broader message from the launch is that the bottleneck in AI-assisted software work has shifted. Teams can generate more code with the help of agents, but moving those changes safely into production remains constrained by access controls, system complexity and governance requirements.

That concern has become more acute as companies test agent-based tools in live environments. Without clear controls, engineering leaders risk giving automated systems broad access to production resources without enough context to make safe decisions or enough oversight to ensure accountability afterward.

Akuity is positioning its platform as a governance layer for that transition. The company was founded by the creators of Argo CD and Kargo, two tools used in modern software delivery workflows, and says its platform already manages delivery pipelines across multiple instances and clusters.

One customer cited early operational results from testing the technology in a production setting.

"Within hours of enabling Akuity's Agentic Control Plane, it identified more than one hundred degraded applications, and in under ten minutes, it flagged one project as the source of a systemic issue, traced it to the root cause and had a fix ready to go. That's exactly the workflow we wanted: identify the problem, tell us if it's systemic and hand us the fix. This is already bringing value to the team, and we are looking to expand its adoption," said Michael Goodness, Principal DevOps Engineer, MLB.

Akuity's leadership framed the launch as a response to the growing role of AI agents in production software operations rather than only in development. The company argues that the challenge is no longer whether agents will be used, but whether businesses can control how they interact with critical systems.

"AI agents have moved past writing code to operating how infrastructure changes get delivered and promoted to production, whether we've built the guardrails for it or not. The winners in this next phase won't be whoever adopts AI fastest, it'll be whoever gives their agents a safe, reliable path to do that work. That's what Akuity's Agentic Control Plane is for," said Wang.