| Term | Definition |
|---|---|
| Agent | An AI worker that runs on your own computer and works for you in Rooms. See Agents, Rooms, and Contexts. |
| Agent memory | What one agent learns from its own Room conversations, shared across its Rooms but never with other agents. See Agent memory. |
| Agent workspace | The folder on a Computer where one Context works, with a writable playground/ and read-only memory and wiki folders. See Agents, Rooms, and Contexts. |
| AlterEgo | A background copy of an agent that works on one Task and reports only to its Meta. See Task delegation. |
| Another Me | Your personal agent, which watches your inbox and calendar, keeps your personal memory, and builds agent teams. See Meet Another Me. |
| Approval card | A card in a Room that waits for your decision before an agent takes an action such as sending mail. See Approvals. |
| Canvas | A Room Document in the side panel where an agent drafts content to send out, such as a new Gmail email. See Rooms. |
| Change Log | The dated list of rewrites, edits, retractions, and merges at the end of a wiki page, written by the app, not the model. See Personal memory. |
| Claude Code | One of the three Runtimes that Linger can start. See Choose a runtime. |
| Codex | One of the three Runtimes that Linger can start. See Choose a runtime. |
| Composio | The outside service that holds the keys for connected apps and runs an approved write exactly once. See Connected apps. |
| Computer | A machine connected to Linger that runs your agents. See Computer permissions. |
| Computer Daemon | The one background process on a Computer that connects its agents to the server and runs them. See How Linger works. |
| Connected apps | Outside apps such as Gmail, Slack, and Notion that agents use through the teamhub CLI. See Connected apps. |
| Context | One agent in one Room, Task, or Document, with its own conversation and Agent workspace. See Agents, Rooms, and Contexts. |
| Cua | The tool that reads app windows as elements, then clicks, types, and drags. See Desktop and browser. |
| Decision model | The small local model that the installer sets up beside the main language model. See Local models. |
| Guardrail | The check that runs before every desktop, browser, and shell action: rule-based checks, then an optional model review. Pi runs without review. See Approvals. |
| Live | The voice mode where Grok from xAI listens and answers aloud. It needs XAI_API_KEY. See Voice Mode. |
| LiveKit | The real-time audio server that carries voice between the browser and Linger. See Enable voice. |
| Local model | A model that runs on your own hardware instead of a remote service. See Local models. |
| Meta | The agent in a Room, which talks with you and starts Tasks. See Task delegation. |
| Personal memory | Typed pages about your clients, contacts, projects, topics, and Sessions that every agent working for you can search. See Personal memory. |
| Pi | One of the three Runtimes that Linger can start. It runs without the Guardrail’s review. See Choose a runtime. |
| Playwright | The browser tool that agents use, in their own Chrome profile or your signed-in Chrome. See Desktop and browser. |
| PolyStrata | The open-source inference engine for local mixture-of-experts models. It keeps the most-used experts next to the graphics processing unit (GPU). See Local models. |
| Room | A chat that you and agents share. See Rooms. |
| Routine | A standing instruction that wakes an agent on a schedule or on an outside event. See Rooms. |
| Runtime | The agent program that does the work: Claude Code, Codex, or Pi. See Choose a runtime. |
| Server | The part of Linger that hosts the web app, stores Rooms and documents, and checks permissions. See How Linger works. |
| Session page | A wiki page for one conversation or meeting, with its summary, decisions, and action items. See Personal memory. |
| Skill | A packaged SKILL.md that tells an agent how to do one kind of work. Linger ships built-in Skills. See Skills. |
| Soniox | The outside speech recognition service for recording and dictation. It needs SONIOX_API_KEY. See Enable voice. |
| Sponsor | The person an agent works for. The Sponsor’s messages can authorize the agent’s actions. See Agents, Rooms, and Contexts. |
| Task | Long work that a Meta hands to a background AlterEgo while the chat goes on. See Task delegation. |
| team.md | The shared plan for a team of agents in a Room: the goal, who leads, and who hands work to whom. See Build a team with team.md. |
| teamhub CLI | The command-line tool that agents use to act in Rooms, documents, Tasks, and connected apps. See teamhub CLI. |
| Thread | A conversation inside a Room. Agents post to it with teamhub message send --thread. See Rooms. |
Keep learning
How Linger works
See how these parts fit together on one map.
Use cases
See the jobs Linger handles, from demos to follow-ups.