Computer Agent capabilities

A Computer Agent starts with its published instructions and model. Add capabilities based on the work it is expected to perform.

​

Skills and connected services

Skills teach the Agent how to work with a service or workflow. Some are ready immediately; others require configuration, credentials, a file, or an authorization step. Complete every required setup action and verify the skill before assigning live work. If authorization opens a sign-in page or displays a device code, the account owner may need to participate. See Skills, accounts, and credentials.

​

Browser access and reusable helpers

The Computer Agent has its own browser environment. A user can open temporary interactive access when a website needs sign-in or manual confirmation. For repeatable work, use a reusable helper. A helper can capture a stable browser workflow or perform focused API, file, and data work without opening the browser. Record and generate browser helpers, then test them with representative data. Use broader Computer work when the Agent must reason across several tools, skills, or changing pages. When generating a browser helper, you can give its skill a neutral reusable name. This is useful when the recorded website address contains a customer name or another detail that should not become the skill’s identity. See Reusable helpers and browser automation.

​

Workspace files

The workspace stores customer-authored project files, skill files, reference material, and automation artifacts. Organize files by project and use clear names so the Agent can find the right material. Agent instructions and managed configuration are edited through the Agent editor rather than as ordinary workspace files.

​

Canvas

Canvas is a simple web application the Computer Agent can maintain for dashboards, reports, review queues, and other visual workflows. Edit the source, publish it, and verify the live page. Choose public, private, or off access based on who should be able to view it. See Build an Agent Canvas.

​

Memory

Memory is appropriate for compact facts, preferences, decisions, and recurring context. It is not a place for credentials, complete transcripts, or large documents.

​

Channels

Channels let people interact with a Computer Agent through services such as Slack. Limit access to the intended people or channels, decide whether mentions are required, and test inbound and outbound messages before announcing the connection. See Connect a Computer Agent channel.

​

Credentials and approvals

Use the managed connection or secret setup provided by Vida. Do not place credentials in instructions, chats, memory, or workspace documents. The Agent should ask before monetary, destructive, account-changing, or otherwise sensitive actions unless the user has already established a clear approved policy.