ByteCompute joins Acenta as a launch partner

Inference is part of the experience

Today, we’re announcing ByteCompute as a launch partner of Acenta. Together, we’re collaborating on inference solutions provided by ByteCompute to support the agents people run on Acenta.

An agent needs more than a model. It needs a runtime, tools, files, and a place to work with people. It also needs inference: the service that processes a model request and returns the output the agent uses to continue its work.

Acenta brings the working environment together in an Agent Box. ByteCompute contributes inference solutions behind managed model access. Our collaboration connects those two parts of the experience, making it easier for users to put agents to work without configuring the underlying inference service themselves.

What the collaboration means for users

The focus is practical: make model inference available as part of a managed agent environment. When an Acenta agent uses a ByteCompute-backed model, Acenta sends the model request to ByteCompute and returns the result to the agent. The user works with the agent through OpenAgents Workspace.

That gives each part a clear role. Acenta manages the Agent Box and its connection to the user’s work. ByteCompute provides the inference service for models routed through its platform. OpenAgents Workspace provides the place to give tasks, share context, and review results.

ByteCompute is one of Acenta’s model providers. The collaboration and joint privacy commitment described here apply to ByteCompute-backed inference through Acenta. Current models and plan details are listed on Acenta’s pricing page.

A shared commitment to protecting user data

Acenta and ByteCompute are committed to protecting the data users entrust to their agents. Prompts may contain project context, working notes, or other information that a user provides to complete a task. That content should be handled for its intended purpose.

For public users, Acenta and ByteCompute omit prompt content from inference logs entirely. When your agent uses ByteCompute-backed inference through Acenta, your prompts are processed to provide the model response, but are not recorded or stored as prompt logs by either platform.

Acenta keeps metering records, such as token counts, the model used, and request timing, to account for usage. Those records do not contain your prompts. Conversations, agent sessions, and files you keep in Workspace or your Agent Box are separate from inference logs and remain part of your ongoing work.

For enterprise customers, we plan to offer usage auditing only when requested, with its scope agreed as part of enterprise support. It will not be enabled by default for public users.

Building useful agents together

The goal of this collaboration is to make inference a dependable part of the Agent Box experience. Users should be able to focus on the task they want done, the context they choose to share, and the result they need to review.

We’re pleased to launch Acenta with ByteCompute and to work together on the inference solutions behind that experience.

Learn more about ByteCompute, or read our introduction to Acenta.