Notion Just Announced A Massive Shift (Big Update)

Notion just announced a massive shift, and I'm incredibly excited because this was a direction I was hoping for. I mean, you and I both know that Notion was already great, but this really pushes things to another level.

The announcement came as part of Notion's first-ever Dev Day. But you don't need to be a developer to be really excited about this and use this. So, in this video, let's go over everything that happened, what it means for your workspace, and how you can make the most of it.

Notion announced three big new features: workers, an agent API, and an agent SDK. Now, that sounds really technical, but again, let's break it all down because you'll notice that all of these things are incredibly useful, even if you never want to touch a line of code.

Before we dive into the details, a quick word on availability and pricing. The whole workers block, right? So, workers, webhooks, and sync is available starting today. I'll show you in a second how to activate it in your workspace so you can start playing around with it. And then the other two parts, right, the agent SDK and the external API, those are still in alpha and beta on a waitlist. So, you still need to sign up for those to get access to them, but you can do so with the links down below in the description. And Notion is rolling out access to flows as we're speaking.

To activate workers in your workspace, you need to be a workspace owner and you need to be at least on the Business plan, right? So, just like AI, workers are available on Business and above. And then you can go to the workers section here under features. And then if you haven't activated them, you'll see a button to activate them. Once you have them activated, this is where you see the workers deployed in your workspace and by whom. You can also decide who can create workers, while you can turn them all off if you want to, and get more analytics for them, just like they're getting them currently with Notion agents further down the line.

In terms of pricing, Notion workers will be a paid add-on because they're using compute, right? And if you run a lot of workers, you use more of the infrastructure. So, it makes sense that based on your usage, this will be priced. But the way that they are priced, right, is magnitudes cheaper than agents. And that's because AI and reasoning is really, really expensive, right? It takes a ridiculous amount of compute for AI to run. Whereas code is really cheap to execute. So while this is paid and will use the same credit system as the credits for agents, you'll spend like a fraction of these credits for it, right?

Notion shared some sample calculations. Basically, if you're pulling your Salesforce tickets, right, every 15 minutes, which is, you know, very, very frequent, that amounts to 86 cents per month. If you pull in from Jira once per day, right, that's 1 cent. And even if you have some heavy usage, right? Okay, 9,000 agent runs per month, right? That calls your Zendesk tickets. Um, that's still only like $13, right? So it's a fraction of the cost. Plus, they are free until August 11th, right? So until August 11th, we can deploy and build as many workers. I expect that just like with Notion AI and Notion agents, right, we'll see analytics plenty ahead of time so we can optimize them. Um, but yeah, we get charged per run, and runs are much, much cheaper than agent executions.

First up, workers. Workers were actually announced by Notion a few weeks ago already. Um, but they were in super early alpha, flew a bit under the radar. Now though, it's the time to really figure out what they do. In particular, you know, if you're non-technical, it can sound a bit intimidating. What does a worker do, right? But actually, it's fairly straightforward. Basically, a worker is a tiny bit of code that you deploy to Notion servers, and then it can do things inside of your Notion or outside of it. You can think of it as if, you know, you're kind of giving your Notion AI or your Notion system in general access to a new thing that it can't do out of the box. And yes, it requires code, but actually, in the age of AI, it's very easy to create that code yourself. We're actually going to release a video later this week with sort of like a complete base tutorial, right, of even if you have no idea how to, you know, touch your terminal and do this, right, how you can create your own worker. So make sure to subscribe. But just try to get the concept, that's the important part.

You can again think of it like this specialized tool that you give it, or like a certain automation, right? Uh, you might have in Notion already played around with Notion automations. This like little flash icon might. You can say, "Okay, if a project moves to the status, then do something." Or you might have played around with outside automation tools like, you know, Make and Zapier. Your workers are basically like that, just on steroids. Uh, they can fill pretty much any gap that is remaining, right? So anything that you where you currently say, "Hey, my Notion automations, I wish they could do X," well, a worker will be able to do it. For example, Notion automations, right, one thing that they can't do is search. You can't say, "Okay, if something happens in this database, search through my other database, find, you know, where one record equals this value, and then set it." For example, if you say, "Hey, whenever one of our tasks, right, the type is changed to a bug, check in our engineering team who's responsible for bugs and assign that person." With a worker, that's possible. With a normal automation, it's not. So, it allows you, right, to close that gap.

Workers mostly are able to show up in your Notion workspace in three ways. First, you can create a worker and give that worker as a tool to a custom agent. And you can think of that, you know, like creating a power upgrade or, you know, if you play Pokémon, as if you're teaching your custom agent a new trick that it can't do out of the box. So, for example, right, custom agents, they might not have a certain connector. For example, there's no connector for a custom agent to connect to ConvertKit or Kit, uh, the newsletter platform that I use. So, what did I do? I created a small worker, and this worker can access the API from Kit. And now I can give this worker tool to my agent. And now my agent can do that. So when the agent runs, it can pull automatically data from Kit and then use that in its ongoing system. That's one way to use it.

Another way to use them, and this is new, right? This is sort of like announced today. Your workers can be triggered by a webhook, which means you don't need to use AI for this, right? You can just say, "Hey, here's a worker. If X happens, I want you to do this." There are a lot of examples where this comes in handy. We're going to talk about most of them in a second. But just as a quick one, you could, for example, say, "Hey, okay, every time we get, right, this new bug request, check who's responsible for that and assign them." So, basically, whenever something happens in your workspace, do it. And the cool thing is, right, Notion has webhooks built-in. So as an automation on your database, you could say, "Hey, if X happens, you know, send a webhook," and with that webhook, trigger the worker. So, basically, you can now inside of Notion build a complete automation loop.

And last but not least, we have sync. And sync is kind of like a schedule, right? A run job technically, but basically means, "Okay, every 30 minutes, I want you to do something." We're going to talk about sync in more detail in a second, but those are the three main ways, right, that you can invoke or trigger a worker: when an agent decides that it should use it, when a webhook sort of like an alert triggers it, or when, you know, sync schedule sets in.

So, let's take a look at six examples for workers because again, right, this will just make it a bit more practical. The first one, very simple but super useful, a company enricher flow where you can basically say, "Hey, okay, whenever I have something in my CRM, I want to enrich this with additional information." And this can be very simple. For example, there are really a few cool services that give you brand logos for a domain. So you can build a small worker that says, "Hey, whenever we add a new company and fill out the domain, please go fetch the logo and set it as the page icon for this Notion page." Right? It's a small visual tweak, but makes a big difference whether Google now has the Google icon, right, or it's all the same generic one. And then you can, of course, build on this and you could also say, "Okay, don't just grab the icon, also grab data." For example, you could integrate with something like Firecrawl, right, a service that allows you to scrape a company website and use AI to extract key data. And you could then use that, right, to fill out things like, "Hey, what's the industry of the company? What are the latest elements?" Right? Basically, whatever you need to build these auto-enriching flows with a worker. No problem.

Second, task generator. This is one of the most common flows that we used to build with Make or any external tool. The situation is very typical. Um, you might have certain projects that always require the same tasks, right? At least a certain checklist. For example, a new marketing campaign, right? Always has these seven tasks. Uh, and you want to have them autocreate when your project creates. There are a few ways to do this, right? You could add a button, you could add a Notion automation, but most of these traditional ways, they're very ineffective and they don't really scale well. Particularly, it's a mess if you need to update and maintain it. So the gold standard here is to have a separate database with your processes where you define, "Hey, these are the tasks that should happen if, you know, project A type gets generated." And then in the past, we would use an external automation that says, "Okay, you know, new project type A, let's look up in the process list what tasks should be created here and create them." Works very well, but again, you need a separate tool. Now, with a worker, this can happen all inside of Notion where the worker simply, you know, also is triggered via that webhook and now goes in checks that, creates it. Much smoother than before, in particular for companies and larger teams, right, that need always approval process for new tools and might be hard to get something authorized like Make or Zapier. And this is an immense help.

Duplicate checker. Another really cool one. Checking for duplicates in Notion is not a good experience. There's no built-in functionality. And yes, to a certain degree, you could use AI, but it gets expensive very quickly, particularly now that you pay for tokens in code. However, checking for duplicates is somewhat a solved problem, right? It's much more straightforward, at least when the duplicates are, you know, actual 100% duplicates rather than sort of like fuzzy ones where it's like, "Okay, one is Google, one is, you know, Google Ireland," etc. Even that isn't too big of an issue in code. So again, you can create a small worker. Every time a new entry gets added, check against the rest, and if there is a duplicate, you know, like notify us and, you know, or help us merge this.

Overdue escalator. This is more like, "Okay, in regular cadences, let's check what tasks are overdue, right? See how they relate, maybe to certain tracks that we have set, and then notify a person."

External status sync. Right here's where it comes in that a worker can again connect to external tools. So you could say, "Okay, here are our sort of like clients, right? Here are our invoices. Let's check Stripe, right? And see whether this invoice has been paid."

Or one of my favorite ones, um, smarter way to sync meeting notes. If you use Notion's internal meeting notes, but an external tool, you probably know the problem. Um, you want to maybe in Notion already prepare the agenda on upcoming meetings. So, you know, you have in your Notion meetings database already entries for that before the meeting happens. Now the meeting does happen, and you are pulling in the meeting notes. Ideally, you want to not create them as a new entry, but you want to merge them into the existing one. That's something that can be quite tricky, but with workers, it gets a lot easier because here is where you can create actually an agent and give them a tool, and you can say, "Hey agent, once a day, I want you to use this tool to pull in new meeting notes." And the code is going to handle the obvious matching, right, where the titles are identical or the start times are identical. And then for the remaining meetings, right, where code has problems doing this matching, you, as the AI, can go in and do use your smart reasoning and your logic to figure out, "Okay, you know, where does this fit in? Which one needs to be connected?" And then there can be a second tool in the end that again pushes the meeting notes to it. And the reason that we want to use code for these other two steps is that it makes your AI call overall a lot cheaper, right? If your agent had to do all that on its own through reasoning, you would spend tokens for every single bit for the agent figuring out how to pull the meeting notes, for the agent copying these things over there, right? Those are a lot of tokens to burn through. But by giving it a worker, 80% of the task happen on code, which is much, much cheaper, right? The cost to execute code is near zero, whereas the cost to execute an agent, like the reasoning step, that is quite substantial. So that's the other angle, right, that you want to think about when using workers is that they can make your agent runs much cheaper.

Now let's talk about sync. Sync is a subset of a worker. So this is a specific worker type that you can create, just like the one that you add as a tool might for an agent, or the one that gets triggered based on a webhook. And basically, what sync does, it allows you to one-way pull information into Notion. Let's say, for example, you have an external data source, for example, Salesforce is the source of truth for your deal flow, but you want to see those deals in Notion. Now, there's plenty of ways of how you can achieve this, right? You can use a no-code automation. You can use a worker, right? A normal worker to do this, or you can set up sync.

Sync. What sync specifically does is that it runs on a schedule, right? So in a regular frequency, it's going to pull certain things in, and it creates and manages its own database. And that's very important. That means you can't just take one of your existing databases right now and point sync at it. The worker needs to set up its own new database. And then as part of the worker setup, you declare a certain number of properties that get synced from the outside source. So, for example, right, the DNA, the stage, and the value. And those properties are kind of protected and locked down. You can't change them in the Notion UI, and you can't modify the values in it, right? It's a one-way sync. So, you can't in Notion say, "Actually, we want this." Right? For this one-way sync, you would need to update it in Salesforce. But what you then can do is you can add your own Notion properties on top of that in the UI, and those are just completely normal configurable properties. So, you can assign it internally on the Notion side to an owner, right? Can link it in Notion to other elements. And this is a very interesting design because what it solves is this problem that you often have that, you know, data gets pulled in, and then someone changes a property, right, and everything breaks, or someone modifies values, but they didn't actually get modified in the source of truth. So you have the scenario, right, source of truth outside of Notion, I just want to see the data, and then, you know, further link it internally. This is the perfect way to go.

If it's different, right? If you want to build a two-way where you are able to modify things on both ends, you're still better off by building this, you know, either with no-code or with pure code and workers. Plus, there might be also situations, right, where you want to just type this into an existing database. And again, then you can simply use a regular worker. It's not, this is the only way, right, to pull information in, but it just helps you to sort of like gate the information that gets added to Notion.

Here's an example of how this looks like in Notion, right? I used it to pull in my newsletters from Kit. And as you can see, I, while I can open this entry, right, and sort of like have a normal page with properties, I can't click into them. Right? All of these properties are kind of locked for me. I can copy the value, but I can't change the ID, I can't change when it was sent, because all these values, right, they are managed by the agent. Interestingly, I can rename properties, right? And this doesn't break it. So, like there's an ID on the Notion side, but I can't change type, right? I can't switch this away from a text to a different type of property. I also get this little sync thing here that tells me, "Okay, this is part of a worker." And then if I click into the properties, right, we see, "Okay, when I want to edit properties, only the new properties that I created, right, those are the ones I can modify." And here I can freely change anything, right? For test, this can be a select now, right? Full flexibility here. And same for my relation that I created. But none of the others are editable.

One more quick note here, just because this was kind of like a trap that I fell in the beginning when I saw workers when they first came out, right, and before they now got these new abilities, particularly with the webhook triggers. I thought, "Okay, what's the point? Everything that a worker can do, I can do today by building a no-code flow." And this is sort of the, "If you have a hammer, everything looks like a nail." It is correct, right? If you look at a Venn diagram and look at, "Okay, what is the use for no-code automations in Notion and what's the use for a worker?" The overlap is pretty much 100%. Not exactly, because workers can also be called inside of an agent more easily, and that's quite hard to do with the no-code automations. But they fill a similar problem, right? They fill the problem that you need to do more with your Notion tool than the simple things that Notion gives you directly with automations or external integrations can do, and allow you to expand Notion. And this update, that's not going to change, right? It's often the case, "Hey, do I use no-code or do I use a worker?" But I think more and more we should use workers for these things. They're not perfect yet. They have a lot of limitations, like the maximum runtime. Again, all of this in our detailed tutorial later this week. But the general principle holds true. First, code gets easier and easier to create, right? Even if you're not an engineer, thanks to cloud code, Codex, etc., you will be able to create a worker simply by describing what it needs to do. Second, you now have the option to run this inside of Notion, basically, which means one less tool to maintain, one less tool to pay for, and one less tool to get approved from your company. And third, right, and definitely not last, the idea of no-code was always to make it, you know, possible for non-technical users to build these more advanced logic flows. Now, you can do this without it, right? You don't need this middle layer anymore. You can go straight to the source. So, that's something that we already observe, right? Internally, we stopped building no-code tools for our own flows. We're still deploying them, right, in some areas for clients simply because they still have, you know, like areas where they are better than a worker. A worker has a limited runtime, handoffs between workers are limited, right? So often times you still might want to go there, but you see the general direction. I honestly doubt that in six months we're going to still be building using Make or Zapier, and simply because it will be more efficient to go straight to the source. So that's why workers are also exciting because they allow you to, you know, very easily venture into this world. Again, full tutorial coming later this week, but just wanted to, you know, explain this that yes, no-code and a worker mostly do the same thing, but the worker is kind of the upgraded version of it.

So much for workers. Now, let's talk about the agent SDK. Complex word, but don't worry. Again, we're going to break down what this actually means for you. In essence, the agent SDK will allow external tools to talk to your Notion agents, which means yes, it will finally be possible to ask Notion AI a question if you're not in Notion. As you know, there's currently one flow where you can already do this. Notion's custom agents integrate with Slack. So, you can simply message them in Slack, they will respond. But it's A) limited to Slack, and B) it's limited to custom agents. The SDK will work both for custom agents and your personal agent, the one in the bottom right corner, right? That has all your context, the one that has all your information and your master prompt, and just, you know, like a lot more personality usually than a custom one. And that means that, for example, you can be in Claude Co-work, you know, chipping away maybe at a marketing campaign, and you can be like, "Hey, Notion AI, can you pull the latest information from a few meetings for me?" And then Claude Co-work will ping your Notion agent. The Notion agent will respond to Claude Co-work, and you can all do it in one interface without ever leaving this. This is a huge benefit because while tools like J Co-work or Codex are amazing at, you know, plugging into dozens of different tools and orchestrating agent flows, when it comes to working inside of Notion, they are and will always be behind the Notion agents because Notion is investing so much, right, into building the infrastructure, into building the retrieval, and teaching the AI inside of Notion exactly how Notion functions. So it's a lot more effective and efficient for Claude to ask a Notion agent to retrieve information and give it that back, rather than it trying to do it itself. Plus, it means you can sort of like chain things, right? If you build a custom agent and if you give this custom agent a worker, for example, right, you can now trigger this from Claude Co-work. Anyone in your organization can, right? Which means again, you're working on this marketing brief, you're working on Claude Co-work because that's where you have the skill to create a PowerPoint presentation, um, and you need to pull information from Notion, but maybe also from your data warehouse that a custom agent has access to. Now, that's no longer a problem. Or on the other hand, right, let's say you are in Claude Co-work or in Codex, you're doing some development work, and you want to pull the latest specs, you want to pull the latest PDS, or after you're done with your work, you want to make it easy to write these things back to Notion. Codex can of course do that itself, but it can also now give that information to a Notion agent that has all the details on how to properly set us up. So that's basically what the SDK enables. Our external tools will be able to ping a Notion agent and ask it a question, and then it's going to work and do the thing for us.

This also moves Notion much closer in the direction of open floor. Uh, if you're in the whole open floor world, right, in this ability, right, basically talk to your agents from your phone simply by sending them a WhatsApp message or an iMessage, um, this becomes now realistic for Notion because either you can have your open floor talk to your Notion agents, right? That's that's one option to simply integrate it with an existing pool stack, or you could have your own custom setup, right, where a WhatsApp message goes to your Notion agent, right, when you interact with it directly. Both of these paths open up now. So, I, I think there will be a lot of options, right, to just make sure there's no longer a border. You don't need to jump into Notion and use Notion AI through the local UI to profit from everything it can do.

One bigger picture element where this is really important is that this helps us move again a bit towards multiplayer. Right now, a lot of AI work happens in single-player mode because as people flock to Claude, Claude Co-work, or Codex, right, to do more and more processes, that's great on the one hand, but all of the sort of like self-contained local, that's the antithesis to collaboration, right? When you create a skill or when you create a work output that should live in a central system, and Notion is that central system, right? Notion is the central system of record to allow all this collaboration. And since it's now possible to call a Notion agent also when you're working outside of it, that makes it just easier for information, right, to flow back into the system, into the right place.

Now onto the last component, the agent API. You can think of the agent API as the counterpart to the SDK. The SDK allows external tools to talk to Notion AI. So you can talk Notion AI outside of Notion. The agent API allows you to talk to external agents inside of Notion. Right now, right, you already have your Notion agents flying around in Notion doing things, and you see them all in the sidebar. The agent API will allow you to add these external agents basically as first-class citizens. So your Claude agents and your Co-work agents and all your other agents will show up, right, in your agent menu, and you can start interacting with them directly from Notion. And this again, right, closes this gap and allows us to move to multiplayer because as you're working with more and more AI, um, we need the central place where humans and AI can interact. So you can now look basically at a shared Kanban board with all of your tasks and you can say, "Hey, okay, this task, you know, at Claude, please start working on it. This task at Co-work, please start working on it." And this allows you, right, to pick the right agent for the right job. A lot of times, Notion AI will be the best one when you work inside of Notion, right? If inputs and outputs live in Notion, Notion agents are the way to go. To a certain degree, that's also true when some inputs or outputs live outside, right? Notion agents are great when working with calendar, email, and Slack. The other connectors are not as built out yet. So here, Claude has an edge. For example, you're not going to send your Notion agent to actually work on your codebase. But now, right, you can have your central task database where all your engineering tickets go in, all your other items go in, and you can then have your description there. And rather than having to copy-paste this and go to Claude, right, to work on it, you can simply do it from there, and Claude can then check off this task, right, and move it along in the Kanban board, all from this one central interface. So it's basically the agent API, right? The missing piece to truly open up Notion and turn into this, you know, middle orchestration layer for all your services.

And the beginning of the video, right, I said that I'm very excited for this because it's what I was hoping for would happen all along. For the longest time, we always said that Notion is kind of this hub and spoke model. As you grow as a company, Notion will not be your all-in-one tool. You do need specialized tools. And now in 2026, you need specialized agents for certain jobs. But you also need a hub for your company, right? One place where all that information flows together. And by Notion saying, "Hey, you know, we don't care where you work ultimately. You can work with external agents. You can work with internal agents. You can, you know, work from it from any place, but it can all flow back together through here. So it gets routed to the right places, and we all have this one shared record." That's a really, really exciting vision to build too, right? And that makes me so excited to see what else we can do with Notion over the next weeks and months.

Now, how do you interact with all of these things? With workers, with the external agents, internal agents? How do you set this up? Again, we'll do a full tutorial and deep dive into this later this week that you can follow along. But just as a bit of a preview, the answer is basically a CLI, a command-line interface. This is basically a way for you to open up your terminal and through your terminal, log into Notion and authorize yourself, and then through that, you can deploy workers, check on statuses, do all that stuff. Again, if you have never worked before with something like that, don't worry, we'll have a full tutorial that walks you through everything. And even as a non-technical user, you'll be able to use this, particularly thanks to the help of Claude and Codex. But yeah, the CLI is basically what allows you to do everything. NTN is the command, right? So if you know what you're doing, right, you can start running this video today. We link to the official documentation down below. And then, yeah, this is sort of like your one-stop shop to do. This is also the only way to do this, right? You cannot deploy a worker without the command line. You cannot set up all the other things without that. But again, tutorial coming later this week. Make sure to subscribe to the YouTube channel. We'll break it all down. And again, it's fairly simple and straightforward once you understand just a few principles.

So, what does all of this mean for Notion as a platform? Well, basically, Notion is turning into a system of record, a central system that can plug into any kind of AI setup that you need on top, which is amazing because it means you can build your company operating system, have a place for all the knowledge, context, tasks, and then whoever wins the AI race at any given moment, well, you can use that to amplify the power. We all know at this point, right, that context is the most important part to make AI work. And Notion basically turned now into your context in a box that you can take and, you know, give to whatever agent you need. Are Notion's own custom agents the best tool for the job? Perfect. Very easy to set up. Do you need Claude, Codex, or another specialized agent that lives outside of it? Well, no problem. Simply give it access to Notion, and now all of them can interact.

But what excites me even more than just this is the overall direction because it means that Notion is really opening up the platform. It's not locked in where the only way, right, to get something is to wait for the team to ship it. And that's kind of the core Notion spirit. If you were watching my videos in the very beginning, you know that the first two or three years on this channel were pretty much all about figuring out, you know, like fun little workarounds and you are duct-taping Notion features together to do things that were not really built in yet, right? And as Notion keeps closing more and more feature gaps, that's great because now we can do things out of the box, but it also means we start thinking bigger, right? How cool would it be if we could have this or that? Well, Notion is basically handing us the keys, right? The ability to execute code inside of Notion. These things mean that we're going to be able to build pretty much anything, right? Anything in Notion. That also means that there's never been a better time to start using Notion or to push your setup to the next level.

If you need help with that, let me know. I'm on a mission to build Europe's number one Notion consultancy. We want to build at the very edge of AI and figure out how do you design organizations where humans and AI work side by side. I don't think the future is AI taking all of our jobs and just doing everything on its own. I think the future is about AI and human collaboration. So, if you need help navigating these crazy changes and get your team, you know, AI-native, or you're just looking to get started with Notion and really make sure you have the perfect setup, then well, let's talk. Full details for that in the description.

And if this got you all excited about AI in Notion, but you don't know where to get started, well, then check this out next. It's our full tutorial on Notion skills. And AI skills are pretty much the single most important thing that you can learn in 2026. They're the building block for everything. So just click here, and I will see you in the next one.