Vecbase Cloud launch

Vecbase Cloud is the international main product surface for the Vecbase company operating system thesis: one workspace should be able to hold the agents, files, tools, runtime, usage, and permission boundaries that a working company needs.

This release note records what is now part of that product surface. It is intentionally concrete. Research claims live in Research; product narrative lives in Blog; this page is for the shipping record.

1 workspace

Operating boundary

Agents, sessions, files, usage, billing, and access control are grouped around the workspace.

N agents

Execution layer

A workspace can hold multiple agents with different roles, model choices, histories, and permissions.

1 bill

Cost surface

Model usage, credits, subscription state, and top-ups are visible as workspace operating cost.

What shipped

Workspace-centered agent console

The main app is organized around workspaces and agents instead of a single personal chat timeline.

Agent creation and management

Users can create agents, switch between them, inspect settings, choose model behavior, and continue sessions.

Visible tool execution

Search, file reads, writes, commands, sub-agent calls, and other tool events render as visible work cards in the session.

Shared Drive and file output

Workspace files, generated artifacts, previews, uploads, and downloads now sit in a shared file surface.

Sandboxed work runtime

Agent work runs through isolated runtime infrastructure with scoped credentials and workspace storage access.

Billing and usage controls

Subscription state, top-ups, usage visibility, and hard balance enforcement are part of the workspace lifecycle.

Why this release matters

The product direction is to move from "assistant next to a user" to "agents inside a company workspace." That requires more than a better prompt box. The agent needs files, tools, permissions, a place to run, and a cost boundary.

Release line

Now

Workspace as the product primitive

The workspace becomes the boundary for agents, files, sessions, runtime resources, and cost.

Now

Agent work with durable traces

Tool calls and generated artifacts remain inspectable inside the session and workspace.

Repeatable operations

Upcoming releases will deepen schedules, skills, channels, review flows, and workspace memory.

Operator guidance

Use one workspace for one real company, project, or client boundary.
Create agents with narrow jobs rather than one vague generalist.
Store reusable outputs in Drive instead of leaving them only in a session.
Turn recurring prompts into Skills after the manual version is reliable.
Watch workspace usage as operating cost, not as an abstract token counter.

Next focus

The next product focus is repeatability: schedules, reusable Skills, stronger integrations, clearer review states, and better workspace memory. The goal is to make the agent team easier to trust without hiding what it did.