The words · DMN docs

The workspace

Agent: A coding assistant, like Claude Code or Codex, working in a project.
Pane: A terminal or chat in the workspace that hosts one agent, or a plain shell.
Fleet: The agents working on a project right now. The Fleet panel lists them, with what each is doing.
Workspace: A set of panes, shown as a tab. You can have several, and open more windows.
Agent dock: The panel beside the fleet with the board, the agents' activity and their changes (Ctrl+J).
Omnibox: The one search box for everything: commands, files, code, panes, and handing a task to an agent (Ctrl+P).
Board: The project's cards: to-dos agents post and take, and steps only you can do, such as an install, a sign-in or a decision.
Automation: A run on a schedule, once, or while you're away: an agent prompt or a command. One an agent adds goes live at once, unless agent autonomy leaves arming it to you.
Teammate: An agent with no pane that another agent hands a task to (dmn agent spawn), and collects the result from.
Agent autonomy: Your setting for what agents do beyond their own pane without asking you: Autonomous (the default), Guarded or Manual.
Account: A Claude sign-in DMN can start agents under. New Claude panes use the active one; a running pane keeps the one it started with.

Engine: DMN's background process. It indexes your projects and answers agents' questions. The command line calls it the daemon.
Project: A folder DMN indexes and serves. dmn projects lists them, each with a short name you can use in place of its path.
Index: What DMN knows about a project's code, kept on your machine and updated as files change.
Curated section: A whole function, class or block of code that answers a question, cut to fit a token budget, in place of a file read end to end.
Note: A short markdown file an agent saved in the project's .dmn/memories folder. Every agent on the project can read it. Agents and the command line call notes memory.
Rule: A note you marked kind: rule. Briefings list rules first, before any agent's notes.
Briefing: What an agent is told when a session starts: how to ask DMN, the project's rules, then the newest notes, each with its age and a flag when the code it names has changed since.
Lesson: An error agents ran into and fixed in two or more sessions, with its one-line fix. When an agent's command fails the same way, DMN hands it the fix. The command line calls lessons gotchas.
Claim: An advisory hold an agent puts on files it's about to edit, with an expiry. Other agents can see it; it never blocks a write. Agents call it a lease.

Skill: A file that teaches an agent to use the dmn command line. dmn skill-init installs it.
MCP: The Model Context Protocol: the standard way agents call outside tools. dmn mcp-init adds DMN's MCP server to your agents.
Hook: A command an agent runs at set moments, such as the start of a session. dmn hook-init adds DMN's hooks to Claude Code and Codex.
Proxy: A program on your machine between Claude Code and Anthropic that compresses context requests keep repeating, so it costs fewer tokens. On by default for Claude Code agents started in the app; Settings ▸ Context tiering turns it off. The app calls it context tiering.