I think your actor is so tight niche specific that 99% people reading this have absolutely no idea what is this all about, including me 😆 But wish you luck with it!

Congrats @apify on the x402 🤝 with Coinbase! Having been involved with crypto stuff myself at some point in my career, I know just how hard/complicated it is (from both a dev and regulatory perspectives). This is definitely unlocking a new era! (We're now at agents developing, agents reviewing -on Apify's side- and agents paying & using) Should we retire now? 😆

Wrote a piece this weekend I am pretty happy with, featuring the Apify-built-and-maintained MCP CLI client

mcpc

I'm basically using it for on-demand MCP skills loading to minimize prefill latency when using my coding agent instead of loading them in fully https://lws.io/blog/pi-harness-cold-start/

Question for the team: we ship pay-per-event Actors that agents call via mcp.apify.com (verified end-to-end — search → call-actor → dataset, works great). But our deep-report Actor runs 20–60 min, and chat agents won't sit through that. Is the recommended pattern still submit-then-poll with a second call, or is there something coming (MCP Tasks?) that lets an agent fire-and-collect a long run? Happy to be a test case for whatever's in the pipeline.

cc @jiri.spilka

https://cdn.discordapp.com/attachments/1374720660873678882/1529111468681859103/image.jpg?ex=6a60bf87&is=6a5f6e07&hm=2ea38182c9c64045bcec5aa3024c584507d6a9d5682fcbe76e45cf2b87220488& https://cdn.discordapp.com/attachments/1374720660873678882/1529111469063672010/image.jpg?ex=6a60bf87&is=6a5f6e07&hm=b2f2c098e3bcf244fabb8fdd141e7162e5ec2551cf1b6348fe2963649c789c81& https://cdn.discordapp.com/attachments/1374720660873678882/1529111469453611008/image.jpg?ex=6a60bf87&is=6a5f6e07&hm=7f5f065b2e8f56b6ed8c4375ee8dfe0aad35ca0ac7d758ad5ecc00a65c45f236& https://cdn.discordapp.com/attachments/1374720660873678882/1529111469793214586/image.jpg?ex=6a60bf87&is=6a5f6e07&hm=0b900647ce995025b794771205b5e0a0b5b0942283bdd2f034129dddd33d7826&

hey @lukaskrivka eva from support sent me your way for a billing question i've got an actor calling apify/openrouter internally, so my users don't need their own llm key. the part i can't pin down: when my actor kicks off apify/openrouter mid-run, whose account eats the openrouter cost? the user who ran my actor, or mine? i get the injected APIFY_TOKEN is the run owner's, but couldn't find where it says which account the child actor bills to. kind of the dealbreaker before i build this out. thanks!

When you use APIFY_TOKEN env var, that is owned by the user and thus the open router runs under the user account and is billed to the user outside of your Actor. You can pass your own Apify token to the OpenRouter, then it starts under your account and is billed to you. We are figuring a way for this to work as part of PPE billing where it would be subtracted from payouts of the primary Actor rather than separately billed but it is still pretty far from done.

perfect, that settles it. injected token = user pays, my own = i pay. i'll stick with injected. good to know the PPE-integrated version is on the way too, will keep an eye out. thanks Lukas!

I saw the recent x402 implementation. I wanted to know does VAT play a role or how is it calculated?

I can't create GitHub Copilot in any way. Can someone please help me? Thank you sir. Assalamualaikum.

Thank for the great information 🗝️

Published an MCP server Actor last week. Sharing one design choice in case it helps anyone building the same shape. commodus67/sports-probabilities-mcp runs in Standby and exposes three Monte Carlo sports models — NFL, MLB and 17 soccer leagues — as five tools: team probabilities, league table, model vs market, what-if scenario, and archived season history. The server contains no simulation code at all. Every tool call ends in Actor.call() against one of my three already published Actors, cached in a named key-value store (3h for NFL and MLB, 6h for soccer) with an in-memory layer and a stampede guard on top. If the model lived in two places the numbers would drift, and an agent would eventually report a different probability than the Store run the user paid for. Two things that bit me, both Standby-specific. First, in Standby the run belongs to the caller, so my child Actor.call() runs were billed to the user on top of my event price — fixed with a dedicated scoped owner token, and I see from the thread above that this is a known rough edge. Worth noting for anyone hunting for it: the environment variable section is at the bottom of Source > Code, not in Settings, and re-editing a secret blanks the value and the Secret toggle. Second, testing. Streamable HTTP, stateless, on POST /mcp. I added a holdSeconds input that keeps the web server alive during an ordinary run, so I can hit /mcp through the run's Live view URL without a token. Every MCP bug I caught before launch, I caught that way. https://apify.com/commodus67/sports-probabilities-mcp

Update on the MCP server Actor I posted about here at the start of the month: it now covers five sports, NFL, MLB, NHL, NBA and 17 soccer leagues. The decision that made the expansion cheap is the one I described back then: the server doesn't reimplement any of the models. It calls the published Actors with Actor.call and caches the result, so there is exactly one source of truth and the MCP can't disagree with what the Actors already return. Adding NHL and NBA was two entries in a leagues map, not two new Monte Carlo engines. Five tools: get_team_probabilities, get_league_probabilities, compare_model_vs_market (live Kalshi prices), run_scenario, get_probability_history. https://apify.com/commodus67/sports-probabilities-mcp

🗓️ Schedules are now in the Apify MCP server 🎉 Tell your agent "run my task insta-daily every day at 9am Prague time" and it creates a real Apify schedule. Works for Actors and saved tasks. You can pause, re-time and delete them too. Add

?tools=schedules

to your mcp.apify.com URL to switch it on.

Hey everyone! 👋 I built a new Apify Actor for scraping Skool: 🎯 Skool Leads & Pricing Extractor It extracts community names, creator details, member stats, and pricing tiers (Free vs Paid). 🔗 Actor link: https://apify.com/armin_s2001/my-actor Would love your feedback or any feature suggestions! https://cdn.discordapp.com/attachments/1374720660873678882/1552295453700268052/dataset_my-actor_2026-09-21_09-26-47-413.xlsx?ex=6ab5174c&is=6ab3c5cc&hm=dda6a3412c6d57dd226c8edcfd6f90072319bb9ac35108b0de7ce156165ef5d8&

Hey everyone! 👋 I built a new Apify Actor for scraping Skool: 🎯 Skool Leads & Pricing Extractor It extracts community names, creator details, member stats, and pricing tiers (Free vs Paid). 🔗 Actor link: https://apify.com/armin_s2001/my-actor Would love your feedback or any feature suggestions! https://cdn.discordapp.com/attachments/1374720660873678882/1552349288078647378/dataset_my-actor_2026-09-21_09-26-47-413.xlsx?ex=6ab5496f&is=6ab3f7ef&hm=939e2100d032cedd4b5ecdc4ffc79a3ff563a675acf09089b99af8e59a47799e&

Building an outbound list in two runs, and the first step of each is free Which companies? Company Search filters by country, industry, headcount, founded year, HQ city, and funding round, amount and date. Turn on previewOnly and it only counts: German companies with 20–200 people that raised since 2024 come to 626, and seeing that costs nothing. Rows are $1.50 per 1,000. If the database holds a company twice, the repeat comes back as a free "duplicate" row, not a second charge https://apify.com/b2bsearch/company-database-search Who works there? Paste up to 1,000 of those domains into People Search and run mode "market" first. You get titles, seniority, countries and top employers for free. Then take only the people you want at $0.95 per 1,000, charged per person, never per search page. You can cap people per company, skip everyone from last week's dataset, and match "cto" as a whole word so directors and doctors stay out. https://apify.com/b2bsearch/people-database-search There's no login and no cookies, and nothing is scraped live: 1,000 companies come back in about ten seconds. Both work from Claude or Cursor through Apify MCP. We're new here, so tell us which filter you'd want next Just send it to your Claude and ask a request for the data and see how it finds what you need

Hi all 👋 I've wrapped my Actors in an open-source MCP server, so Claude, ChatGPT or Cursor can call them as simple tools (search_flights, search_hotels, get_youtube_transcripts, find_local_businesses and 13 more). Each call runs the Actor on the user's own Apify account, with a spending cap per run. https://github.com/retracn/automationnation-mcp Also on Smithery as automationnation/data-tools. Feedback on the tool design or which Actors you'd want next is very welcome.

message has been deleted

Hi everyone 👋 I built a pay-per-event Sina Weibo & BlackCat Intelligence Scraper for China market research, brand sentiment tracking, and AI training corpora. Most existing Weibo scrapers require your personal login cookies for creator timelines, hit an anonymous pagination wall at ~25 posts, or charge $35/1k on 4GB RAM. What this Actor covers ($3.80 / 1k items, runs on 256MB RAM): - **9 modes with zero cookies / no login**: Keyword search, realtime stream, creator timelines (

user_posts

via dual PC/Mobile fallback), Hot Search Top 50, post comments with English province mapping (

广东

->

Guangdong

), KOL profiles, and Sina BlackCat consumer complaints. - **Unbiased linguistic slicing**: Bypasses Weibo's 30-post anonymous search cap using 80+ neutral Chinese grammatical particles so sentiment distribution isn't skewed. - **Built-in enrichment**: Auto EN->ZH brand translation (e.g.,

Tesla

->

特斯拉

,

Luckin

->

瑞幸咖啡

), full-text expansion (

>140

chars), and zero-RAM lexicon sentiment scoring. https://apify.com/axon-studio/weibo-scraper

If you connect Apify's MCP server to Claude/Cursor, we just published 7 tools on official public data (EU tenders, SAM.gov, NPI Registry, UK Companies House, Mexico business directories). One URL adds them all: https://mcp.apify.com?tools=hexadatalabs/eu-tenders-ted-scraper,hexadatalabs/sam-gov-opportunities-scraper,hexadatalabs/us-healthcare-providers-npi-scraper,hexadatalabs/uk-companies-house-scraper — every description starts with "Use this when…" so the agent picks the right one. Happy to hear which sources you'd want next.

console.log("🚀 My first public project: Ethical World, powered by Apify's open-source Crawlee and AI 😌🤖");https://github.com/patrikdjasnik-eng/ethical-world- im vibec - coder 😏

Shipped **Hexadata Tools**: an Apify Actor that is a remote MCP server (Streamable HTTP) with 15 deterministic business validation tools for AI agents — tax IDs of 15 countries, EU VAT (VIES), IBAN, CLABE/ABA/BIC, LEI, postal codes, phones, FX rates, business days, HS/NAICS codes, VAT rates, Mexican SAT CFDI catalogs and payroll. No LLM inside, pay per call, invalid input free. Also listed in the official MCP registry as io.github.4robot/hexadata-tools. https://apify.com/hexadatalabs/business-validation-tools-mcp — feedback welcome.

For anyone whose agent does SEO research through mcp.apify.com: two tools that work well as plain MCP calls, no login or cookies. - Keyword volume (

plainrows/keyword-search-volume

): Google volume, CPC, difficulty, intent and 12-month history for up to 10k keywords, 94 countries. $0.006 to $0.01 per keyword with data; keywords without data are never charged. Optional AI search volume. - Backlinks (

plainrows/backlink-checker

): backlinks or referring domains of a domain/URL, with a "new links only" mode. Tip for agent builders in general: give the agent a hard budget per call (

maxTotalChargeUsd

) and a small first batch, then let it decide whether to expand. It keeps the bill predictable when the agent loops. Disclosure: both are mine. Questions or field requests welcome.

Hey everyone This looks quite interesting for open-source AI builders https://x.com/svonava/status/2117664970762846449