Once you merge, all Actors are automatically deployed => Does "all Actors" here literally mean all actors or all actors that were changed?

Only Actors that were changed. For releases that includes stuff like README, Changelog (which are skipped for tests). Each Actor defines its "context". By default, that is its Docker context minus

.dockerignore

. So you can structure the repo as you wish: Some Actors can be completely isolated in their folder, some can point context to root (shared code) and some might only depend on some folders

(e.g. /lib/)

. In that case you can use

overrideActorContext

instead of relying on Docker context to specify which changes should rebuild that Actor. Btw you don't have to use the 'release' workflow if you want to deploy manually, you can just choose only the test of PR update and scheduled tests. But we found out deploy on merge very nice.

After apify store category automatically pickup changes , Many of tools not get single user . Any suggestions for keep traffic on tool via Apify store page.. Thanks

Thanks Lukas — this is timely. I maintain a growing fleet of public Store actors and have been wanting a lighter way to rebuild/test only what changed. Going to try apify-test-tools on a few of mine and will share what breaks / what helps.

@avabuildsdata_30321 just advanced to level 1! Thanks for your contributions! 🎉

Looking forward to your feedback! The change detection is one of the main features. We classify changes into 3 groups: - Functional - affects Actor's behavior, needs testing - Cosmetic - only affect how the Actor looks (Readme, schema titles, etc.). For example, we look at changes in schemas, and if you only change the title or description, we don't retest. Also if you merge in master to a branch, we can analyze if it interacts with current code and decide to test or not - Ignored - like .gitignore, internal docs etc.

Will the "All Actors" option scale to a hundred or a thousand actors? Say for a shared lib change. In my experience, the apify api doesn't scale past a dozen or so parallel pushes.

Currently, it would eat all your memory and then crash since we count to have enough memory buffer on the testing account. We can definitely add a queue for builds and tests. But having to go through 1000 Actors will take a long time. Not sure if testing every single Actor is a good idea then, especially if they are based on the same base. Tell me more about your setup, and we will accommodate. The queue is a no-brainer.

Question for the team about where paying demand actually comes from, with my numbers attached in case they are useful to other new creators. I published two Actors in late October (ATS / job-board data, pay-per-event). The supply side is working: 653 company career sites resolved across nine ATS platforms, nightly health checks, zero dead boards. Reaching anyone who pays is not working: 28 Store page views in a month, two users total and one in the last 30 days, revenue $0. Two things I would genuinely like to know: 1. Is there a queue of customer Actor requests with a budget attached that creators can pick up? I know about Apify Experts and the "request an Actor" form, but not whether a creator can see incoming requests and build against a real buyer instead of guessing what the market wants. 2. How is Store search ranked - by users and runs, or by title and description match? Concrete case: my Actor sits at #19 for "job listings api" but is absent from the top 100 for "ats jobs api" and "career site jobs", even though both phrases are literally in its title. If ranking is mostly popularity-weighted, then a new Actor cannot be found through Store search at all, and the honest advice to new creators would be "bring your own traffic", which is a very different plan from "publish and iterate on the listing". Not asking anyone to promote my Actor. I am trying to work out whether the Store is a distribution channel for a new creator or a storefront that only converts traffic you bring yourself, because those two answers lead to completely different work.

So we now have apify ai that build custom actor so developer is needed no more lol

Most customers have their personal agents and if you tell Claude Opus 5.5 to create an Actor from scratch, it will do a pretty good job. So this isn't changing anything. Customers mostly want to know that someone is taking care of the Actor. It is not about making first version but maintenance, communication, pricing.

Hi, quick question, If I switch my existing account, which has public pay-per-event Actors with paying users, to the Creator plan (1$ monthly), will my Actors stay public and keep earning PPE revenue and payouts as normal? Im running out of 5$ month pretty quick so thought this is a way to go and develop … thanks

hi, my message promoting my apify insights mcp github repo was deleted. could you let me know if it violated a rule or how i should share it properly? thanks!

hey all, i built an mcp server for apify creators like me and open-sourced it. connect it to claude, cursor, codex or gemini, then ask about your own actors. it tells you which ones lose money and the price where each would break even. it also shows which runs are failing, which actors get views but few runs, and what apify's quality score wants you to fix. it's read-only and runs on your machine. revenue and profit need a small chrome extension, which only talks to the apify console and your computer. https://github.com/Devansh-365/apify-insights-mcp setup is two commands. if something breaks or you want a tool added, open an issue or ping me here.

Are you going to train this or use the code from non apify developer codebases? If yes that wouldn’t be a good thing to the devs who have strategic ideas on how their scraper works.

I don't think they would do that considering already maybe up to 90% of actors are AI generated so that would be training on AI itself 😆 BTW if you want to keep your code safe you can use apify just as a frontend to your own hosted solutions. I've been doing it with all actors with solutions AI hasn't come up to yourself (yet)

So the compute happens on your servers?

Yes. You basically create own API you can connect to anything, including APIfy 😉

That would hurt if you have a lot of free users right? 50% of my users are free users.

Depends on what do you mean by that. If you take your own hardware as fixed costs it doesn't matter too much how many users are free. if your infrastructure can handle it.... and moreover you can easily limit free users usage. Sure Apify was created to exactly avoid necessity of having own hardware but there are some edge use cases (multithreaded/multicore scraping and parsing, multiple scripts running at once, database support etc.) when apify itself wouldn't work. And also if you want to feel more comfortable regarding crucial parts of your know how. And last but not least you can connect your solution to other api markets.

also, your private infra is usually less expensive than apifys

We don't read your code, and we don't train anything. We only build harnesses on top of common models. We have over 10 years of experience building large-scale Actors to build such harnesses (e.g., "how to analyze and search traffic on the website", "skill to deobfuscate JS" etc.), so we don't really need your code for this either. Moreover, frontier models got so good that it can already implement pretty much everything, building Actors is moving closer to knowing what users want.

What do you mean by Apify not working for multicore scraping? Actor RAM goes up to 32 GB which should be about 8 CPU cores. If you care about performance and cost reduction, the new Rust Crawlee/SDK can be up to 7x faster and cheaper on compute (+ I added some API reduction optimizations); see https://blog.apify.com/crawlee-rust. If you combine that with multitenant server mode so all your users overlap traffic, I think you can get cost on Apify very close to $0 for pretty much any load. Like having 100 concurrent requests on single 1 GB run.. Unless you are using browsers, but you can always reverse engineer the target to some extend.

Thx, will definitely have a look at this

What is multi tenant server mode?

new feature that is being released in approximately 24 minutes from now 😄

Give your Actors a new way to be used 🚀 @User Today we’re launching Server Actors (formerly Actor Standby mode): a way to host your Actors as services that receive HTTP requests and return results directly. With multi-tenant mode, your Actor runs in your account and serves requests from multiple callers through a shared pool of instances. What you get as a creator: - Extra visibility in Apify Store with a dedicated Server Actors collection and filter - New use cases for your Actor and new users who come with them - New tools to build by combining several Actors into one service - Lower cost on single request as you can share a run between different users The input interface is still a good fit for bulk collection and large crawls. You can offer either interface or both, depending on what works best for your Actor. Read docs to enable server interface on an existing Actor or build a new one: http://docs.apify.com/actors/running/server-actors Check out Server Actors created by Apify & Community Beta testers: https://apify.com/store/collections/server-actors

I will test it out

Good luck with your talk at Run by Apify today, @johnisleaping_24391 ! 🚀🎤 📣 https://run.apify.com

Thank you!