Machine interface
For agents
You operate X Stack. Live JSON is at x-stacks.vercel.app. If you found this UI on xstack.grok.me, still fetch from Origin — same-origin catalog JSON is a hard 404 (not a seed, not a 301). First hop:
llms.txt → index.json → status.json. Do not invent.Rules
- JSON origin is https://x-stacks.vercel.app. Same-origin grok.me catalog-shaped JSON must 404.
- generated_at is last_success (scrape). published_at on this UI is the bundle deploy, not a catalog refresh.
- Call it freshly scraped only if last_run_status is ok and now is before last_run_fresh_until (last_run + 6h). Frozen ok after that is overdue.
- Grok Build versions: catalog.channels and x.ai/cli/stable, alpha, enterprise. Do not use product.urls.changelog.
- can is a short verb. notes is markdown for humans. summary is the short factual line. Do not invent rows.
You operate X Stack. Fetch https://x-stacks.vercel.app/llms.txt then index.json and status.json. Watch loop: watch.json and playbook grok-bot-watch-catalog plus /updates/{filter}.json. Same-origin https://xstack.grok.me/catalog.json must 404 — not the catalog. generated_at is last_success; a UI published_at is not a scrape. Freshly scraped only if last_run_status is ok AND now is before last_run_fresh_until (last_run + 6h). Grok Build versions come from catalog.channels and https://x.ai/cli/stable|alpha|enterprise — not x.ai/build/changelog. can is a short verb; notes is markdown for humans. Do not invent. If a source is 403 or product freshness is stale, say it was not freshly scraped. Return the answer contract.Endpoints
- GET https://x-stacks.vercel.app/index.jsonSlim first hop + freshness
- GET https://x-stacks.vercel.app/status.jsonScrape health (ok | 403 | skipped)
- GET https://x-stacks.vercel.app/watch.jsonNamed subscribe slices
- GET https://x-stacks.vercel.app/updates.jsonNew changelog ids at this generated_at
- GET https://x-stacks.vercel.app/updates/imagine.jsonWatch slice: Imagine
- GET https://x-stacks.vercel.app/updates/bot.jsonWatch slice: Grok Bot
- GET https://x-stacks.vercel.app/updates/build-alpha.jsonWatch slice: Grok Build alpha
- GET https://x-stacks.vercel.app/updates/voice.jsonWatch slice: Voice
- GET https://x-stacks.vercel.app/updates/cursor.jsonWatch slice: Cursor
- GET https://x-stacks.vercel.app/updates/x-api.jsonWatch slice: X API
- GET https://x-stacks.vercel.app/catalog.jsonFull catalog
- GET https://x-stacks.vercel.app/playbooks.jsonPlaybooks
- GET https://x-stacks.vercel.app/changelog.jsonFlat changelog
- GET https://x-stacks.vercel.app/features.jsonFlat feature index
- GET https://x-stacks.vercel.app/llms.txtOrigin contract, plain text
- GET https://x-stacks.vercel.app/.well-known/llms.txtSame contract
- GET https://x-stacks.vercel.app/catalog.mdMarkdown dump
generated_at (remote) · X Stack
2 sources marked 403/skipped — not freshly scraped.
Answer contract
Every agent reply should be this shape. missing is required when the JSON does not support part of the question. Use playbook_id from playbooks.json.
{
"playbook_id": "string | null",
"product_ids": ["..."],
"answer": "short colleague-style briefing",
"sources": [{ "url": "https://..." }],
"confidence": "high" | "medium" | "low",
"missing": []
}Example questions
- Which always-on product fits this job? (always-on-route)
- Watch this catalog and message me when it changes (grok-bot-watch-catalog)
- Current Grok Build alpha and enterprise channel versions
- Features tagged computer-use
- Changelog for Grok Bot and Cursor Cloud Agents since 2026-08-01
Update loop
Origin owns the catalog at https://x-stacks.vercel.app. Watch: watch.json, then updates.json or /updates/{filter}.json (playbook grok-bot-watch-catalog). If a source is 403 or product freshness is stale, say it was not freshly scraped. Same-origin catalog JSON on this UI host is a hard 404, not the catalog.