Agents · Checked on 2026-10-04
DMN for OpenCode
Use OpenCode next to Claude Code or Codex without starting it from zero. Through DMN's skill it reads the notes the other agents saved, searches the same local index of your code, and sees which files they're editing.
What OpenCode gets
- Notes from every agent, saved in your project's
.dmn/memories. - Code search in one call. The matching code comes back, not a list of files to read.
- Callers and blast radius. Who calls a function, and what a change to it could break.
- Claims on files. It sees which files other agents are editing.
What's different from Claude Code
Claude Code and Codex get a briefing hook, so each session starts with your rules and the newest notes. OpenCode has no such hook from DMN yet: it reads the notes when it asks for them, through the skill. It runs in any terminal, a DMN shell pane included.
Set it up
OpenCode reads the skill DMN installs for Claude Code and Codex, so if you've connected either, there's nothing more to do. For OpenCode alone:
dmn skill-init --apply --agent opencode
Leave out --apply to see where it would write first. To check or undo it, see
Connect your agents.
DMN for Windows is coming soon. The commands work once it's installed.
Questions
Does DMN use my model provider's account?
OpenCode keeps using your own providers, as it does without DMN. DMN doesn't sell AI.
Does my code leave my machine?
Not through DMN, unless you point it at a remote embedding service yourself. The engine only listens on your own machine (127.0.0.1), and the index and notes stay there.
Will OpenCode use fewer tokens?
We haven't measured it. The benchmarks ran Claude Code; other agents may gain more or less.