Docs

npm i -g @avr00/clawd-html
export CLAWD_TOKEN=clawd_live_…          # from Settings → API tokens
cat index.html | clawd publish --title "My page" --visibility unlisted

The shareable URL is printed to stdout, so it composes: URL=$(clawd publish page.html).

Create a token in Settings → API tokens. It's shown once. Tokens are tied to your account and grant the same access you have in the app. Revoke anytime from the same screen.

Resolution order, highest priority first:

1. --token flag
2. CLAWD_TOKEN env var
3. ~/.config/clawd/config.json (written by clawd login)

clawd login clawd_live_…     # verifies, then saves to config
# or, non-interactively:
export CLAWD_TOKEN=clawd_live_…

publish options#

# from a file
clawd publish index.html --title "Pitch" --visibility public

# straight from a model, into a workspace, expiring in 7 days
some-llm | clawd publish -t "Draft" -w ateam -e 7d

# round-trip
clawd get aj3p2k9q > saved.html

A folder groups pages and is shared at its own URL. Opening a folder you're allowed to see grants access to every page inside it — even individually private pages — so one link shares the whole set. A personal folder holds your personal pages; a workspace folder holds that workspace's pages (a page adopts its folder's scope).

clawd folder create "Weekly analyses" --visibility unlisted
cat week-1.html | clawd publish --title "Week 1" --folder weekly-analyses
# share the printed /f/<slug> link — recipients can open every page in it

Base URL https://clawd-html-puce.vercel.app. Authenticate with Authorization: Bearer clawd_live_…. All responses are JSON.

curl -X POST https://clawd-html-puce.vercel.app/api/v1/pages \
  -H "Authorization: Bearer clawd_live_…" \
  -H "Content-Type: application/json" \
  -d '{"title":"Hello","content":"<h1>hi</h1>","visibility":"unlisted"}'

# → { "id": "...", "slug": "aj3p2k9q",
#     "url": "https://clawd-html-puce.vercel.app/v/aj3p2k9q",
#     "visibility": "unlisted" }

POST body fields: content (required HTML string), title, visibility, workspaceSlug, expiresInHours.

private and a workspace are mutually exclusive. Expiry is optional; an expired link returns a friendly "expired" screen and stops being viewable.

Every page and folder gets a live "watching now" pill in its header. Anyone viewing — signed in or not — shows up in the count. Only the page owner (and workspace members on workspace pages) can see other viewers' names; everyone else just sees the number.

The manage page for a page or folder also shows Audience: a list of signed-in viewers with timestamps, plus a single muted line totaling anonymous viewers from the last 7 days. We don't store IPs, user agents, or any fingerprint for anonymous traffic — just a count.

• Human messages go to stderr; the shareable URL (or --json payload) goes to stdout.
• --json on every read command for structured parsing.
• Non-zero exit code on any error, message on stderr.
• File or stdin input — composes in any pipeline.

We publish a skill for the skills.sh ecosystem at our /.well-known/agent-skills/ endpoint. One command installs it into Claude Code, Cursor, Codex, and 50+ other agents.

npx skills add https://clawd-html-puce.vercel.app

Installs the use-skills skill, an opinionated, pitfall-focused guide to the npx skills CLI itself — source formats, scope (-g vs project), cross-agent installs, common errors. Your agent activates it automatically when you mention skills.

We expose a Streamable HTTP Model Context Protocol server at https://clawd-html-puce.vercel.app/mcp. Auth is the same API token as the CLI — pass it as a Bearer header.

claude mcp add clawd https://clawd-html-puce.vercel.app/mcp \
  --header "Authorization: Bearer clawd_live_…"

{
  "mcpServers": {
    "clawd": {
      "url": "https://clawd-html-puce.vercel.app/mcp",
      "headers": { "Authorization": "Bearer clawd_live_…" }
    }
  }
}

Tools exposed:

A clawd page can be a single HTML string or a multi-file bundle: a directory of HTML/CSS/JS files that link to each other with relative paths. The bundle is served at /v/<slug>/ and sub-paths so <a href="page-two.html"> just works. Entry point is index.html when present, otherwise the first .html file.

# A directory of linked HTML files
clawd publish ./design-system/
# → https://clawd-html-puce.vercel.app/v/aj3p2k9q

# Sub-paths resolve to siblings
# https://clawd-html-puce.vercel.app/v/aj3p2k9q/page-two.html

clawd files aj3p2k9q             # list what's in the bundle
clawd update aj3p2k9q ./design-system/   # replace the whole bundle

curl -X POST https://clawd-html-puce.vercel.app/api/v1/pages \
  -H "Authorization: Bearer clawd_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "title": "Design system",
    "visibility": "unlisted",
    "files": [
      { "path": "index.html",    "content": "<a href=\"button.html\">Button</a>" },
      { "path": "button.html",   "content": "<h1>Button</h1>" },
      { "path": "styles.css",    "content": "body{font-family:system-ui}" }
    ]
  }'

Limits: 64 files max, 400 KB per file, 1 MB total. Allowed extensions: .html .htm .css .js .mjs .svg .json .txt. Binary assets (images, fonts) aren't supported yet — reference them by external URL.