Loom — AI-assisted workflow for VS Code

Your AI coding agent forgets everything every session. Loom is the memory it doesn't have.

Loom turns your project into a document-driven loop between you and your AI — chat → idea → design → req → plan → implement → done. Every stage is a Markdown doc the AI reads, writes, and tracks across sessions, so it never starts empty and never loses the thread.

🔗 Get Loom: Marketplace · Open VSX · CLI on npm · GitHub

📚 Guides: Which way is for you? · Core concepts · Extension guide · CLI / Claude Code

Install — one click, no CLI

1. Install this extension from the Marketplace (search Loom AI, publisher reslava).
2. Open a folder and click Initialize Loom in the Loom panel.

That's it. The Loom engine is bundled inside the extension and runs on VS Code's own runtime — no npm install, no global command, no Node to set up. Initializing creates .loom/ (config + the CLAUDE.md AI session contract), loom/ (your doc workspace), and .mcp.json (so an AI agent can drive Loom too). Update the extension and your session rules refresh themselves — nothing to re-run.

📚 New here? → install to your first idea in five minutes.

The loop

1. Chat — open a chat doc, think out loud with the AI (AI Reply).
2. Idea — Create Idea, or Promote a chat straight into one.
3. Design — Generate Design defines how to build it.
4. Plan — Generate Plan breaks it into concrete, reviewable steps.
5. Implement — Do Step: the AI implements the next step, records a done-note, marks it ✅.

Each stage is a Markdown document in the Loom panel. Nothing disappears between sessions.

The Loom panel

The Loom Activity Bar icon opens two views: Threads and Context.

Threads + Context — every weave → thread → doc, each node carrying its derived state: locked reqs, plan step counts (3/3 · done), staleness ⚠. The Context panel shows the exact docs — and their token cost — the AI will receive, before you launch.	Roadmap — a Show Roadmap toggle re-lays the tree into the derived cross-weave roadmap: present + future in dependency-then-priority order, over a history of shipped versions. Drag-to-reorder writes soft priority. All computed from the documents.

_{Click either image to view full size.}

Reports — synthesize a project overview, architecture, decisions, or release notes from the doc graph, right from the panel; saved under loom/reports/.

The buttons

AI actions run the agent:

Create / promote (Create Idea / Design / Plan / Chat, Promote, Add References…) and manage (Start / Close Plan, Set Status, Rename / Archive / Delete, Validate) actions appear inline on each node and in its right-click menu — no AI required.

AI setup — pick one path (only one is required)

• Claude Code CLI (default). If claude is on your PATH, buttons open a persistent Loom AI terminal and run the agent — it reads your Loom docs, calls the right MCP tools, and writes the result back. No API key, works with a Claude Pro subscription.

  npm install -g @anthropic-ai/claude-code   # the recommended path — free with Claude Pro
  

• API key (fallback). No CLI? Set reslava-loom.ai.apiKey (and .provider) in VS Code settings and the extension calls the provider directly.

Settings

Requirements

• VS Code 1.85+ — self-contained (bundled server; no separate CLI or Node install)
• For AI actions: Claude Code (recommended — no API key) or an API key in settings

Also runs headless

Prefer the terminal, or driving Loom from Cursor / Continue / CI? The same engine ships as a CLI + MCP server → @reslava/loom on npm.

Telemetry & feedback

Loom can send anonymous, content-free usage telemetry so we can see whether the loop is used and where people stall — off by default, one-click status-bar toggle, first enable shows exactly what's sent (never content, titles, paths, or PII; events go to PostHog EU). A Feedback button in the status bar opens a prefilled GitHub issue you edit before sending — opt-in, nothing automatic. → full detail in the User Guide.

License

MIT