Akamai has unveiled its AI inference service, Akamai Inference, and an isolated execution environment for AI agents, Akamai Agent Sandbox, on its cloud, Techzine reported on Oct. 28 local time.

The report said the two services aim to help companies build, run and protect AI applications and agents on a single platform.

Akamai Inference provides managed access to open-weight and custom models through an OpenAI-compatible API. Developers do not need to rewrite existing applications.

Customers can choose a pay-as-you-go serverless option or a dedicated deployment option. At launch, it offers 12 open-weight models ranging from 8 billion to 280 billion parameters. They include Nvidia Nemotron3 Ultra, OpenAI gpt-oss-120b and Google Gemma 4.

Agent Sandbox runs code written directly by AI agents in a separate space isolated from the agent. The sandbox is based on the Fermyon Spin framework, which Akamai acquired.

It limits computation, memory, external connections and system privileges through WebAssembly-based isolation. Akamai also plans to add functions to control what agents can access and what can access agents.

Adam Karon (애덤 캐런), Akamai's chief operating officer for cloud, said, “We have pursued an AI infrastructure strategy on the belief that AI inference will account for most corporate interest and investment.”