Glossary

The words used across the apps, the dashboard and these docs, in alphabetical order.

TermMeaning
AgentThe WECOMMANDER program on a machine that carries out tool calls. It connects out to the gateway and enforces the device's limits.
Allowed foldersThe only folders an agent's file tools may touch (also called allowed roots).
Allowed commands (allowlist)Command patterns that run without approval under standard trust; process_exec only.
API access keyA secret for MCP clients and scripts that use a bearer token; it acts with the role it was given.
ApprovalA person's decision that one exact high-risk action may run once.
Audit logThe record of every tool call, approval and permission change.
Autonomous modeFull trust for many devices for a time, with automatic expiry.
Classic modeThe Windows app without the Dock, working in windows.
ConnectorThe WECOMMANDER app on your computer when it carries AI calls to SSH servers (Hybrid Access).
DeviceA machine with the agent, enrolled in a workspace.
DockThe taskbar-like bar of the Windows app on a screen edge, with live server meters and panels.
GatewayThe WECOMMANDER service at wecommander.app: MCP endpoint, sign-in, approvals, audit and the dashboard.
Host / serverA saved connection in the desktop app (SSH, Telnet, serial or Mosh).
Hybrid AccessSharing SSH servers with AI through your own computer, with nothing installed on the server.
IdentityA reusable user name with a key or password for many servers.
Known hostsThe servers' host keys you trusted.
Management CenterThe part of the desktop app that handles sign-in, approvals, autonomous mode and the agent; on Linux, the app itself.
MCPModel Context Protocol, the open standard AI assistants use to call tools.
PresetHow a shared SSH server may be used by AI: read only, with approval, or timed autonomous (local only).
Run asWhether an agent's commands run as its normal user or as root.
Safety netA list of catastrophic commands refused on every trust level.
Scheduled taskA command WECOMMANDER runs for you on a schedule.
Secret maskingReplacing passwords, tokens and keys in output with [REDACTED] before the AI sees it.
SnippetA saved command you send into a terminal.
TargetA place AI tools can act on: an agent device or a shared SSH server (targets_list).
Team libraryServers and snippets shared with the workspace, end-to-end encrypted.
Trust levelstrict, standard or full: how many high-risk actions need an approval on a device or SSH server.
WorkspaceThe account unit that owns devices, members, approvals, audit and the plan.