Follow ZDNET: Add us as a preferred source on Google.

ZDNET’s key takeaways

• Tasklet is billed as a chatbot-style AI agent authoring, hosting, and deployment tool.
• A knack for interfacing with any system, API or not, could be its unique superpower.
• Once your integrations are in place, Tasklet can build UIs on top of them in minutes.

By now, most people I know — friends, family, and colleagues — have experienced AI in some way, and many are generally impressed with its utility. There’s also a healthy amount of skepticism to go along with those positive impressions. As skeptics go, I’m a harder customer to please than most. 

I’ve mentioned Tasklet.ai before. In one article about how Microsoft was turning Entra into part of the organizational agentic AI control plane, I cited Tasklet’s existence as evidence that AI agent authoring and deployment would one day be child’s play.

Now that I’ve had sufficient time to put Tasklet to use, I can honestly say that, in my 30 years as a tech journalist, I’ve never been wowed by a technology the way Tasklet has wowed me. 

My first Tasklet agent

Tasklet is hardly just an agentic AI authoring, hosting, and deployment platform. In fact, once you head down the path of creating your first agent, you’ll begin to wonder about the official definition of an AI agent. Perhaps it’s my background as a developer, but after creating my first Tasklet agent, I wasn’t sure whether what I’d built was an AI agent or just software any programmer might write. With one big difference: I didn’t write a stitch of code. 

Also: How to build better AI agents for your business – without creating trust issues

Using natural language, as shown in the video below, I simply told Tasklet what I wanted my first AI agent to do and where to find the data, and it went off and took care of business (cleverly prompting me for input when it needed it). 

The result of my first effort is an agent that notifies me when it’s time to reapply wax to my bike chain. For cyclists, wax is an alternative to traditional chain lubricant. If you’re a cyclist like me who rides more than 5,000 miles per year, you know how easy it is to ride too far (usually over multiple rides) without giving your bicycle’s drivetrain the care it needs. My new chain lube agent not only notifies me (over email) when I’ve ridden 125 miles since the last waxing, it also lets me know when it’s time to have the chain professionally cleaned and rewaxed at my local bike shop.

Here’s what’s extraordinary about how Tasklet built my agent: I only had to tell it where to find my cycling mileage log (an online service called Strava) and which of my bikes to watch for elapsed mileage (my Specialized Tarmac SL6 road bike). It not only built the code to keep track of my miles and send me an email when it was time to rewax or service my chain, but also handled all the integration code to tap into my cycling log on Strava.com. 

Also: How Claude Code’s new auto mode prevents AI coding disasters – without slowing you down

For comparison’s sake, I asked Anthropic’s Claude to do the same thing, and it complained that Strava didn’t have a Model Context Protocol (MCP) interface. It gave me a workaround that wasn’t nearly as slick and frictionless as Tasklet’s approach. Incidentally, if you need an agent to send emails or texts to someone other than yourself, Tasklet will do that too. But, for anti-spam reasons, users other than yourself must opt in to receive those messages.

On the surface, all of this may sound trivial. I assure you that it’s not. 

The no-code dream comes true

For the last decade, under the moniker of “no-code software,” the biggest tech titans, funded by billions of dollars, have been racing to be the first to deliver this sort of utility: a platform that writes software and automagically handles whatever integrations are necessary to make that software work. As the former editor in chief of ProgrammableWeb (once the official journal of the API economy), I’ve seen countless no- and low-code tools that attempted to deliver on this promise, but never did. 

Now, here’s Tasklet making that dream come true. 

Also: I built an iOS app in just two days with just my voice – and it was electrifying

“If you just tell Tasklet where the data is, it will automatically figure out how to get that data and bring it into the app, whether there’s an official API (application programming interface) or not,” Tasklet founder Andrew Lee told me. “In fact, I’m not even sure why we need MCP anymore.”

Wait. I spent 15 years covering the API economy, and here was this founder telling me it all could have been a waste of time? Lee wasn’t suggesting a bit of disruption. He was suggesting eliminating a ginormous segment of the technology business! And what about all the work that went into making MCP the OData-like standard API of the AI industry? Is AI that good — especially Tasklet’s AI — that we can just throw all of that away? 

Surely, he must be mistaken. But as I thought about it more, it made perfect sense. If AI should be capable of anything, it should be capable of poking around the border of any network service, discovering all of its interfaces (official or not, for all you cybersecurity pros out there), and then enabling any other software — even software it develops like Tasklet does — to access that data. (For the record, if you watch the above video closely, you’ll see how, with my approval, Strava issues an OAuth token to Tasklet in order to gain access to my cycling log.)

This is when I began to wonder what Tasklet really is. A natural language-driven no-code software solution? A revolutionary new integration layer for mashing up multiple services into a single app? An AI agent authoring, hosting, and deployment tool? Yes, yes, and yes. But wait (you guessed it) — there’s more. 

Also: How to get free AI headshots that don’t look fake – with Nano Banana 2

While I was testing Tasklet, Lee notified me that they had just released a new feature called Instant Apps. It’s one thing for an agent to go off and do something and then build a report or send you a text or an email on some periodic or triggered basis. But what if you wanted a full-blown app that included your own custom-built front end? Tasklet has you covered. 

I’m a big Notion user. In fact, after I ditched Evernote for Notion (I love Evernote, but it just got too expensive), I asked Tasklet to migrate all my Evernote data into Notion. Actually, I didn’t need Tasklet to handle the migration because Notion has its own import utility. But I wanted to see how Tasklet handled mashup scenarios involving two or more disparate online services.

More than an AI agent authoring and deployment tool

After giving Tasklet the necessary permissions to access my Evernote and Notion accounts, the migration went more smoothly than I expected. It’s an example of where Tasklet is more than just an AI agent authoring and deployment tool. It can also handle one-time integration tasks (or ongoing workflows between two or more systems of record). 

But if you’ve ever used Notion, you know that its front-end for entering data into a Notion database is hard to customize. For example, one of my personalized Notion databases tracks my billable time for clients who pay me by the hour. Every time I perform a chunk of work for one of these clients, I clock in and clock out by adding a new record to that database. 

Also: 7 AI coding techniques I use to ship real, reliable products – fast

By default, Notion’s user interface for entering the time of day is pretty clunky. I’m sure it’s fixable if I dig deeply into Notion’s documentation. But after watching Tasklet’s video about Instant Apps, I wondered if I could use Tasklet to build my own front-end — one that made it easier to create most of a new record (including the start and finish times) with mouse clicks instead of the keyboard.

As shown in the screenshot below, I gave Tasklet a very simple natural-language prompt: “Create a front end to my Notion time tracker database that defaults to the current date and time for entering data into the clock-in and clock-out fields.” 

As also shown in the screenshot, Tasklet tried to get me to upgrade from its Advanced tier (which Lee gave me free access to) to the Expert tier, saying that “building an instant app works best at Expert intelligence for cleaner code and fewer build errors.” But I continued without upgrading. 

As shown in the next screenshot, after granting Tasklet permission to access my Notion account, it snooped around, figured out which of my many databases was the exact one to use, and set about building my new front end. 

As shown in the next screenshot, the phrase Instant Apps lived up to its promise, instantly creating the exact front end I asked for. 

On the left side of the screen is a summary of what Tasklet did, and on the right is the Instant App itself. I’m pretty sure I blurted out “Oh my God” the moment the app was created. 

The idea that, with a simple line of natural language (and about 5 minutes of time), I could customize the front end of one of my online accounts left me completely stunned. As shown in the final screenshot below, the Tasklet event also created a mouse-driven user interface for entering the date and time. What’s crazy is that I forgot to ask for that. It just did it anyway. 

I was able to get Claude to build a similar front end to Notion because, unlike Strava, Notion has an MCP interface. But it took multiple tries to get the mouse-driven time picker to work. Meanwhile, Tasklet nailed that part on the first try without even being asked. 

The implications of something like Tasklet — something that frictionlessly handles back-end integrations while cleverly assembling front-end entry and reporting interfaces — are mind-boggling.

Never mind the developer

For decades, one tech industry Holy Grail has been to gift ordinary users the superpowers of software developers. On the reporting side, back in the 1990’s, before web APIs came along, open database connector (ODBC)-enabled reporting tools were one of the early attempts at this idea. Before that, when I was a developer, my clients would come to me with the craziest automation ideas that, in their minds, would make them significantly more productive. Weeks or months (and many dollars) later, I would come back to them with the finished product. 

Also: Why AI is both a curse and a blessing to open-source software – according to developers

But now, if you have a vision for what exactly would make you or your team significantly more productive — something crazy like a front end that gives you input and reporting tools across a bunch of different back ends — Tasklet will more than likely get you to the finish line in less than a day or maybe even an hour. 

And then you can toss some agents into the mix to alert you when certain thresholds are exceeded (e.g., inventory is dangerously low) or to automate whatever other crazy ideas you have. Never mind the developer. 

Oh, and if you need to ask it a regular old chatbot-style question, it can do that too. 

Wow. Just wow.