Skip to content
velora

Documentation

One place to understand how Velora works.

Velora is a desktop workspace for coordinating AI work across cloud models, local runtimes, and supported coding agents. You provide the outcome. Velora turns it into a visible plan, keeps related tasks connected, and pauses supported high-impact computer actions for your approval.

Documentation reflects the current product direction. Exact controls may change before a public build.

Getting started

From an objective to a supervised result.

  1. 1

    Create your account

    Use the same Velora account on the website and in the desktop app. Account access keeps your plan and Pro entitlement tied to you instead of one browser.

  2. 2

    Connect the tools you choose

    Add supported cloud providers, a local runtime such as Ollama, or a compatible coding agent. Velora does not automatically gain access to every model or application on your computer.

  3. 3

    Describe the finished outcome

    Give Velora the goal, useful constraints, and the material it needs. Review the proposed plan before dependent work begins.

  4. 4

    Review the work and approvals

    Follow each task, inspect handoffs, and decide whether a supported action that sends, deletes, publishes, or changes data should run.

Workflow model

Plan, dispatch, approve, complete.

Plan

Velora breaks a broad request into smaller tasks and shows how they depend on one another.

Dispatch

Independent tasks can run in parallel. Dependent work waits for the context it needs.

Approve

Supported high-impact actions can pause with the target and effect shown in plain language.

Complete

Results, decisions, and action events remain attached to the workflow for inspection.

Models and agents

Bring the services that fit your work.

Velora is designed to coordinate supported services from OpenAI, Anthropic, Google, DeepSeek, local Ollama models, and compatible agent command-line tools. Provider availability depends on the credentials, plans, regions, and policies attached to your own accounts. A provider reference does not imply endorsement or a commercial partnership.

Permissions and safety

Access should match the task.

  • Inspection and change are separate. Reading a supported interface does not automatically grant permission to change it.
  • Approval names the target. A request should explain the action and the item it affects before you decide.
  • The action is logged. The workflow records whether an action ran freely, was approved, or was refused.

Availability

Built for macOS and Windows.

macOS

Apple silicon and Intel build details will appear here when signed installers and system requirements are ready.

Preparing release

Windows

Windows 64-bit and ARM64 packages are planned. Download buttons will be enabled only after signed installers are published.

Coming next

Product media

Real product captures will replace every placeholder.

The website currently uses neutral product frames where a final interface capture is not available. These captures will provide enough material to replace every placeholder with the real product.

  • Main workspaceA clean 16:10 screenshot with a real plan, three tasks, and no personal files or API keys visible.
  • Approval requestA close view of a realistic send, delete, or publish action before approval.
  • Model routingA workflow that shows two cloud providers, Ollama, and a coding agent working on connected tasks.
  • Action historyA completed log with freely run, approved, and rejected events.
  • Short product recordingA silent 12 to 18 second 1440p recording that moves from goal, to plan, to approval, to completed result.
Read product updates