The delivered system runs on the Akka Agentic AI Platform, and the customer chooses who operates it. The customer owns and extends it on the platform, or Akka keeps running it under the agreement that delivered it.
Akka hands over the specifications along with the running system. A later change is written against those specifications, so the account of why each part exists arrives with the code.
A forward-deployed engineering engagement runs for months and hands back a bespoke system. The customer then owns code written by people who have moved on, and the reasons behind each part moved on with them.
Akka Specify produces the system by generation from specifications, and those specifications are handed over with it. The account of why each part exists is the specification, versioned in the repository beside the code.
The Akka runtime is BSL-licensed, so a customer keeps the right to run, modify and self-host it. Deployment runs on Akka cloud, inside the customer hyperscaler VPC, or on the customer Kubernetes infrastructure.
The specification describes the system in its own terms, so agents stay portable across deployments and Akka versions.
The system runs on the same platform in both cases, so the move changes who operates it. The specifications, the code and the evidence record belong to the customer throughout.
A monthly corpus watch detects the change and scopes the review to the control rows it affects. A prompt change is a half-day cycle, and a foundation-model swap runs for weeks and is budgeted before it starts.
A defect is reproduced from the interaction record, which captures every run without sampling. The fix is verified by replaying the same events against the changed code.
Extending the system means editing the specification and letting the check verify the change. Developers on the customer team write code against the Akka SDK for the parts they want to write by hand.
Every tier except Sandbox includes 24/7 SRE monitoring, no-downtime rolling updates and fully resilient infrastructure. In-VPC tiers install the platform inside the customer cloud or datacenter.
