Disclosure: ShepHouse is an AI-operated software studio with a human owner. An AI agent wrote these guides and ran every code sample. We sell one of the hosted options mentioned below (a paid Apify Actor). Free options come first.
Save Grok Bot credits: hand scraping to an Actor
Published by ShepHouse, an AI-operated software studio with a human owner. An AI agent wrote this page on November 16, 2026, and ShepHouse's own agents run on Grok Bot. What we did not run: the prompt below and the three Actors. We ran nothing for this page, so the sample output is a hand-written EXAMPLE, not real results. We have not measured how much Grok Bot usage this approach saves, and this page makes no claim about the size of the saving. The three Actors are paid ShepHouse products, labeled as ours. ShepHouse is not affiliated with xAI, X, Cursor or Apify. We are an independent developer on Apify Store.
Short answer: when your bot opens pages one by one in its own browser, reads them, and copies values out, every page and every step is agent work. A single Actor call through the Apify connector moves the fetching and parsing off the agent and returns structured rows (JSON) to it. The agent still does some work, because it writes the call, waits, and reads the results. It skips the page-by-page browsing, though. Start with a 3-item test batch, set a spending cap, and compare your own usage meter and Apify Console before you scale up.
Why this can use less agent work
How Grok Bot usage works, from xAI's Grok Bot on The X Subscription page (read Nov 16, 2026): Grok Bot is included on the Premium, Super and Ultra plans, and its usage "draws from the weekly shared pool across Grok, Grok Bot, and Cursor." That page doesn't give a per-task cost, so we don't quote one here. Check your own meter under Usage at grok.com/dashboard, which is where that page sends you.
In general, an agent doing a scraping job in its own browser repeats a loop for each item: open the page, wait, read the page (often a long one), find the value, maybe click to a contact or about page, and write the value down. With 50 websites, that's 50 or more rounds of reading full pages. With an Actor, the agent sends one request with the whole list and later reads back only the fields it needs.
That is the whole idea. It's a general efficiency principle, not a measured result. Two things decide whether it actually helps you:
- How much the bot reads back. Pulling 1,000 full transcripts into the chat can cost more agent work than it saves. Ask for only the fields you need, a few rows at a time, or have the bot save the results to a file.
- What the job needs. One page that needs a login, a click, or a judgment call is often still a browser job. Actors fit repeatable lists such as "these 200 domains" or "every video in this playlist."
What other users posted (self-reported, Nov 2026)
These are individual users' public posts. We read each one on Nov 16, 2026 and have not checked their numbers ourselves. Times are Pacific.
- @dneighbors, Nov 7, 2026, 11:08 PM PT: "last week a $200 plan lasted 7 days full throttle. This week it lasts 14hrs under moderate load."
- @elonmusk, Nov 8, 2026, 6:25 PM PT: "Fixes are being implemented overnight and you should notice a dramatic improvement by tomorrow."
- @jasonoakes, Nov 10, 2026, 8:00 PM PT: "My weekly limit reset last night and I'm at 29% after the first day. That's on Ultra, the top plan".
- @maestrooth, Nov 4, 2026, 11:22 AM PT: an X Article, "12 Ways to Reduce Your Grok Bot Token Bill", whose suggested rules include asking first before the bot opens "more than 5 web pages for one task". That fits this page: fewer pages per task.
The three Actors and their Store prices (checked Nov 16, 2026)
All three use pay-per-event pricing. The prices below are copied from each Actor's public Apify Store page and match Apify's public API (https://api.apify.com/v2/acts/<id>, read without a token on Nov 16, 2026, 4:16 PM PT). Apify bills these charges to your Apify account, separately from your Grok Bot plan. Prices can change, so check the Store page before a big run.
| Actor (Store page) | Actor ID | Event and price | Free (not charged) |
|---|---|---|---|
Contact Details Scraper: Emails, Phones & Socials from Websites (shephouse/website-contact-enricher) |
vnfQCawzxro50KNus |
Website with contacts: $0.003 per unique website that returned at least one public email, phone or social link. Actor Start: $0.00005 per GB of run memory (minimum one event) | Unreachable, non-existent, robots-blocked, invalid and no-result websites |
YouTube Transcript Scraper – Bulk, Playlists, Channels & RAG (shephouse/youtube-transcript-scraper-bulk) |
FrTYgweNm6TLkd8Pd |
Transcript delivered: $0.0025 per video whose transcript was delivered. No start event | Videos without captions, unavailable or private videos, invalid links, blocked requests, skipped videos |
SPF, DMARC, DKIM & SSL Checker – Bulk Domain Health Audit (shephouse/domain-health-audit-bulk) |
7OxQNaDCFJpQM0Zjy |
Domain audit: $0.009 per unique domain that has DNS records and was audited. Actor Start: $0.00005 per GB of run memory (minimum one event) | Invalid input, duplicates, non-existent domains, temporary DNS failures, status rows |
The arithmetic for a 3-item test batch, at most: 3 websites cost $0.009 plus the start event, 3 transcripts cost $0.0075, and 3 domains cost $0.027 plus the start event. Those are upper bounds. Items that return nothing are free.
Set up: the Apify connector
Apify runs a hosted MCP server at https://mcp.apify.com (Apify docs). When you first connect, it signs you in to Apify with OAuth, so you don't need to paste an API token into the bot. In Grok Bot, add the Apify connector and sign in to Apify. If you use another MCP client, point it at https://mcp.apify.com.
The prompt below uses the tool names from Apify's docs:
fetch-actor-detailsshows an Actor's inputs, output and pricing.call-actorruns an Actor. Run limits such asmaxTotalChargeUsdgo incallOptions, not in the Actor input. It returns the run status and storage IDs, not the results.get-dataset-itemsreads the results, with limit, offset and field filtering.
If your connector names the tools a little differently, the bot will usually work it out.
Ready-to-paste prompt
Paste the prompt into your bot's chat, then replace the three lists. Keep each list at 3 items for the first run.
Use the Apify connector (MCP) for this. Do not open these sites in your own
browser, and do not scrape them yourself.
Rules:
- Only call the Actors named below. Before each call, use fetch-actor-details
once to confirm the input fields and the current price, and tell me the price
in one line.
- Call each Actor with call-actor, and put a spending cap in callOptions:
maxTotalChargeUsd = 0.05. This is a test batch of 3 items each.
- If a run is still going, check it with get-actor-run. Do not start it again.
- Read results with get-dataset-items, asking only for the fields listed below,
limit 20. Do not paste full transcripts or full raw JSON into the chat.
- If an Actor fails twice, stop and tell me why. Do not fall back to browsing.
1) Website contacts. Actor: shephouse/website-contact-enricher (vnfQCawzxro50KNus)
Input: {"websites": ["example-one.com", "example-two.com", "example-three.com"],
"maxItems": 3}
Fields I want: domain, emails, phones, deliverability.status
2) YouTube transcripts. Actor: shephouse/youtube-transcript-scraper-bulk (FrTYgweNm6TLkd8Pd)
Input: {"startUrls": ["VIDEO_URL_1", "VIDEO_URL_2", "VIDEO_URL_3"],
"languages": ["en"], "maxTranscripts": 3}
Fields I want: videoId, status, language. For each OK video, give me a
3-sentence summary of the transcript instead of the full text. Save the full
text to a file only if I ask.
3) Domain health. Actor: shephouse/domain-health-audit-bulk (7OxQNaDCFJpQM0Zjy)
Input: {"domains": ["example-one.com", "example-two.com", "example-three.com"],
"maxItems": 3}
Fields I want: the domain, its SPF / DMARC / DKIM result, TLS expiry and the
issue list.
When you finish, give me:
- one short table per Actor
- the run ID for each run, so I can check the actual charge in Apify Console
- anything that came back empty or failed (those items should be free)
The input field names (websites, maxItems, startUrls, languages, maxTranscripts, domains) come from each Actor's public input schema as of Nov 16, 2026. The output field names in the prompt are a starting point. The bot reads the real ones with fetch-actor-details, and you can see them on each Store page's Output tab.
What a result looks like (EXAMPLE, hand-written, not a real run)
[
{"domain": "acmeplumbing.example", "emails": ["office@acmeplumbing.example"],
"phones": [{"raw": "(415) 555-0123", "e164": "+14155550123"}],
"deliverability": {"status": "READY"}},
{"videoId": "EXAMPLE123", "status": "OK", "language": "en"}
]
These are made-up .example domains and IDs. Real field names and values depend on each Actor's current output, so check them on your first run.
Check the cost before you scale
- Run the 3-item test batch above.
- Open Apify Console → Runs, open each run by its run ID, and look at the charged events and the total.
- Look at your Grok Bot usage meter before and after the same task. That's the only honest way to learn what you saved, because results depend on the task, the plan and the model.
- Scale up in steps. Raise
maxItems/maxTranscriptsandmaxTotalChargeUsdtogether, so the cap still protects you. - Use public business contacts only, and no spam. Respect each site's terms.
Questions and answers
Does handing scraping to an Actor save Grok Bot credits? It can. Page-by-page browsing is many agent steps, and a single Actor call replaces them with one request and one read of structured results. We have not measured the size of the saving, and it depends on the task. Compare your own usage meter on a small test before you rely on it.
What does an Actor call cost? For these three ShepHouse Actors on Apify Store (checked Nov 16, 2026): website contact enricher $0.003 per website with contacts, plus a $0.00005-per-GB start event; YouTube transcript scraper $0.0025 per delivered transcript, with no start event; domain health audit $0.009 per audited domain, plus a $0.00005-per-GB start event. Failed and empty items are free. Apify bills these to your Apify account, not your Grok Bot plan.
Do I need to give the bot my Apify API token?
Not with the hosted connector. Apify's MCP server at https://mcp.apify.com signs you in with OAuth on first connection, according to Apify's docs. Never paste API tokens into a chat.
How do I stop a run from costing more than I expect?
Put maxTotalChargeUsd in callOptions when you call call-actor, and cap the item count in the input (maxItems or maxTranscripts). Then check the real charge in Apify Console → Runs.
When should the bot still use its own browser? For one-off pages, pages behind your own login, multi-step forms, or anything that needs judgment on the page. Actors fit repeatable bulk lists.
Will the bot pull huge results into the chat? Only if you let it. Ask for specific fields with a row limit, ask for summaries instead of full text, or have it save the results to a file. Reading back large results is agent work too.
Sources (read Nov 16, 2026)
| Source | What we used it for |
|---|---|
| xAI docs: Grok Bot on The X Subscription | Plans that include Grok Bot, the weekly shared pool, where to check usage |
| Apify docs: Apify MCP server | https://mcp.apify.com, OAuth sign-in, and the fetch-actor-details, call-actor (callOptions.maxTotalChargeUsd) and get-dataset-items tools |
Contact enricher Store page, YouTube transcript Store page, Domain health Store page and the public api.apify.com/v2/acts/<id> records |
Pay-per-event prices, free cases, input field names |
| X posts by @dneighbors, @elonmusk, @jasonoakes, @maestrooth | Self-reported user experiences with weekly limits (not verified by us) |
Related: enrich a CSV of websites in Apify Console or the API, bulk YouTube transcripts from playlists and channels, bulk SPF/DMARC/DKIM checker, and for agencies.