name

ents-zed

description

Use git-ents comments inside the Zed editor via the ents-zed extension and the ents-lens language server. Use when working in Zed and asked to see, leave, or resolve inline comments, or to set up the editor for the ents comment loop.

ents in Zed

The ents-zed extension (at editors/zed/) registers a language server, ents-lsp, that runs git ents lsp. The server projects the repo's comments (refs/meta/comments/*) into whatever buffer you have open and lets you leave new ones — no MCP, no network, just an LSP over stdio reusing the local composition root.

The language server is the same mechanism the CLI and web UI use (lens.parity), so a comment left in Zed is readable and resolvable from git ents comment and the web, and vice versa. If you are acting on comments programmatically rather than through the editor, use the ents-comments skill (the CLI) instead — it is the same entities.

Setup

1. git ents lsp must be on PATH (build/install the git-ents binary). The extension launches git with ents lsp and speaks LSP over stdio.
2. Install the extension as a dev extension: Zed → command palette → zed: install dev extension → pick editors/zed.
3. Turn on code lenses. Zed renders LSP code lenses but they are off by default. Enable them, or the inline comment lenses won't show:
   • setting: "code_lens": "on", or
   • command palette: editor: toggle code lens.

What renders

• Code lenses at each open comment's projected line (once enabled): the id, a summary, and View / Reply / Resolve actions.
• Hint diagnostics for the same comments — these show inline even without code lenses enabled, which is why the server publishes them (lens.diagnostics). They are hints, never warnings or errors.
• Hover over a commented range shows the full thread.
• Code action on a selection ("leave an ents comment") to compose a new comment; see the crate's compose flow for how saving creates it.

Composing a comment

Trigger the code action on the lines you want to anchor to. It opens a template file (git-commit style: first content is the body, # lines are ignored, an empty body aborts). Saving a non-empty body creates the comment, anchored to the working-tree bytes you selected.

Because everything is the one mechanism, the iteration loop is: leave comments in Zed, then tell an agent to "address all comments in the working tree" (the ents-comments skill) — it reads the same open comments, fixes the code, replies, and resolves them, and your next Zed publish reflects the resolutions.

See editors/zed/README.adoc for the authoritative, version-tracked notes on exactly which surfaces the current Zed renders.