A Spec Kit extension that turns a feature's markdown artifacts into a
human-friendly review surface in the browser. The reviewer reads the rendered
documents, selects text to annotate, and chats comments. Those notes queue up
and, on one click, go back to the agent, which edits the canonical .md files
while the browser live-reloads to show the changes.
- Run
/speckit.axi.review(/speckit-axi-reviewin skills-based agents such as Claude Code). The agent launches a local review surface (Node, zero dependencies, loopback only) for the active feature and opens your browser. - You read the artifacts as web pages. Select any phrase to attach a note, or type a general comment in the chat. Notes collect in a queue.
- Click Send. The agent receives the whole queue and edits the markdown.
- The browser live-reloads. The agent posts a short summary of what changed. Repeat until you end the session.
Markdown stays the source of truth: the agent edits the .md files, never a
derived copy. Annotations and the queue stay local. The renderer (marked,
DOMPurify, mermaid) loads from a pinned CDN, so viewing needs an internet
connection; offline, the surface shows a notice.
The full guide lives in the project wiki. This README is the front door only.
| Wiki page | When to read |
|---|---|
| Home | Overview and reading order. |
| Getting Started | First install, zero to your first browser review in five minutes. |
| Commands | Deep reference for /speckit.axi.review and the review server it drives. |
| Workflow | How a review flows: feature resolution, the annotate and apply loop. |
| Examples | A worked review session and the edits it produces. |
| Annotation Format | The TOON queue the reviewer sends and how each field maps to an edit. |
| Troubleshooting | Server, browser, port, and offline issues, with their fixes. |
| FAQ | Conceptual questions and design rationale. |
| Architecture | How it works at runtime: the technology stack and how information travels. |
The wiki is generated from docs/ on every push to main. To browse
the same content as plain markdown, open the docs folder.
| Command | What it does |
|---|---|
/speckit.axi.review |
Open a feature's markdown in a browser review surface, collect annotations, and apply the notes. |
/speckit.axi.review takes an optional feature directory, for example /speckit.axi.review specs/001-my-feature. Without one, the agent uses the feature that Spec Kit marks as active, then your branch, then the most recent specs/* directory. The command does not generate a new file; it edits the feature's existing markdown in place.
Skills-based agents show the command under another name: /speckit-axi-review in Claude Code, $speckit-axi-review in Codex.
The extension also registers two optional hooks. After /speckit.specify, the after_specify hook offers to open spec.md in the review surface. After /speckit.plan, the after_plan hook offers to open plan.md and the other plan artifacts.
- Spec Kit
>= 0.12.17initialized in the project (specify init). - Node.js 22 or later (the review tool runs with
node, no install step). - A browser and an internet connection for the renderer.
Install directly from the latest release:
specify extension add axi --from https://github.com/d0whc3r/spec-kit-axi/releases/download/v2.0.0/axi-2.0.0.zipChange the version in the URL to pin a different release.
To update, run the same command with a newer release URL and add --force. To see the catalog entry, run specify extension search axi or specify extension info axi. Install by name (specify extension add axi) does not work, because Spec Kit's community catalog is discovery only. See Troubleshooting.
For prerequisites and the first-run walkthrough see Getting Started.
See CONTRIBUTING.md.
Axi was inspired by lavish-axi, and its agent-facing command line is built on the AXI conventions for tools that agents drive over a shell.
MIT. See LICENSE.