The AI Platform is a provider-agnostic AI superapp for multi-agent, multi-human organizations. Work with friends or coworkers, bring your own mini app and share it with the world.
Replies
Best
Maker
📌
This whole journey started from us working on Module Federation, the work we did on federating websites and mobile experiences led one of our customers to ask us a year and a half ago, "Could you federate agents?"
That single question led us to start building, what started with just trying to understand how to work with multiple agents and multiple humans at the same time led to us building a new AI first collaboration platform, that led to us needing to embed a IDE, then a browser, then a workflow engine, then a kanban board.
Eventually we took a step back and looked at the monster that we had created and asked, what is this? Its an AI superapp. Once we wrapped our head around that terminology everything, and I mean everything, changed. Embracing mini apps lead to us being able to take our module federation expertise to the desktop and create an ecosystem. Now every idea, every task, every app we were trying to deploy became lighter, faster, and more easily agentified (if that's even a word) because it's able to leverage the platform as a whole.
We started sharing what we're building with friends in the community and we're excited to share what they've built on our platform.
Report
@zackary_chapple Congrats team🙌. qq when multiple agents and humans collaborate in the same space, how do you prevent agent-to-agent feedback loops or conflicting actions?
Report
Maker
@vikramp7470 good question, every action in our platform has a clear owner, even when there is delegation. We can also delegate both async and sync to attempt to prevent the chaos from a noisy room. Also we keep the conversation visible so its easier to manage.
how do y’all see this from an enterprise perspective? will you be aiming for FedRamp compliance?
Report
Maker
@hamatoyogi we built this enterprise first, so a lot of the controls and complexities (and why it took so long to get to where we are) are because of that. Between being able to decide where your inference comes from and where the compute happens we had to grind through it all.
In terms of FedRamp compliance both AWS and Cloudflare have FedRamp options for inference, and we have the ability to do local inference as well which helps.
For the conversations themselves and FedRamp compatibility we do have BYOC at the larger enterprise levels.
To be honest though I think I'm most excited about the blend of local and cloud inference since it lets me offset the lighter work with local hardware.
One of my favorite things about TAP is that you can build the random little tools you want without turning each one into a whole standalone product.
Miniapps just become part of the workspace alongside the people and agents already there.
I wanted to try that with something pretty far from the usual AI demo, so I adapted Mimaraka’s Daiza into a miniapp.
You drop in a transparent PNG, design an acrylic figure or keychain, preview it in 3D, and export the production files. It all runs locally inside TAP.
It was a fun one to build. Hopefully people bring their own weird, specific ideas too.
Report
Maker
@nsttt need to show this off with some of the fabrication folks
Report
I was actually blown away by how full featured and well built this platform is. I decided to try to make some things with their Mini App SDK, and I think "Mini App" might actually sell what it does short: its full on platform extension with robust integration hooks. I'm excited to experiment with using it to make this app mine, its an absolutely dank feature.
Report
Maker
@ken_wheeler I need screenshots of the Slopometer you shared on the stream! That was pretty interesting
I use The AI Platform to run simulations for students in my courses. The specialists and apps especially are helpful for me to give my students a near-real simulation of what it's like making product decisions. Great product!
Report
Maker
@kentcdodds I can't wait to see the feedback from next week's cohort, there has been a ton of improvements made since the last cohort.
Report
Maker
@kentcdodds I remember day one! It's great to have you on board!
Report
Maker
Stand-ups assume everyone's awake at the same time, and once a team is spread across the Americas, Europe, and Asia that stops being true - someone's always attending at 6am, and the ones who can't make it get reconstructed secondhand. The usual workaround, "post your update async in a channel," lasts about a week before nobody's reading 15 people's updates and the ritual becomes theater.
It's gotten noisier in the agentic era, too. When part of your team is shipping through coding agents, one person's GitHub activity for a single day can be a dozen commits and three PRs - a raw digest of that isn't a standup, it's noise. So Pulse doesn't just dump activity: it drafts your update from it (or you type it fresh), and a specialist rolls the whole team's week into one summary, so a lead gets what actually mattered without wading through the firehose or hosting a meeting to extract it.
Worth calling out separately: we built Pulse as a miniapp on top of the platform, and that's the reason it moved as fast as it did. Roster, tasks, specialists, storage - Pulse doesn't own any of that infrastructure, it just asks the host for scoped grants to use it. That's the concrete answer to "what does building on a platform actually buy you": a full team-facing tool without writing auth, directory sync, or a task board from scratch.
Report
Maker
@billwatson017 leaders in tech are going to have so much more visibility now.
Report
Maker
My team considers me engineering adjacent, lol! I'm on the customer side and I hate CRMs. So I built my own with TAP. I can literally drop in context and it will route accordingly. It's going to be a game changer for me bc now I never have to worry about CRM hygiene. The CRM mini app will do it for me.
You can point any clanker of choice at the Miniapp SDKand make something to share or for yourself. The SDK enables you to create pretty much anything on top of our platform, and makes it agent enabled.
Report
Maker
I got really tired of having to switch apps to look at design files, so I decided to make a design file viewer. A quick miniapp to make reading and editing figma files possible inside of our TAP runtime. Was pretty fun to make and even more fun to see the agents go to town on it.
Report
The AI platform gives you some amazing opportunities through mini apps. As a non-engineer, I was able to build a triage app that brings everything into one place and makes the whole triage process much easier.
Triaging tickets is always fun 😄 But it gets much better when you can keep the entire workflow in one platform: discuss tickets with all the stakeholders, turn those discussions into tasks that your team members or agents can pick up, bring their updates back in, and connect everything into one continuous flow.
For me, that’s one of the most exciting things about the mini app functionality — you can build the workflow you actually need without being an engineer.
Replies
@zackary_chapple Congrats team🙌. qq when multiple agents and humans collaborate in the same space, how do you prevent agent-to-agent feedback loops or conflicting actions?
@vikramp7470 good question, every action in our platform has a clear owner, even when there is delegation. We can also delegate both async and sync to attempt to prevent the chaos from a noisy room. Also we keep the conversation visible so its easier to manage.
@zackary_chapple congrats on the launch to you and your team!
sounds super cool, can’t wait to try it out!
how do y’all see this from an enterprise perspective? will you be aiming for FedRamp compliance?
@hamatoyogi we built this enterprise first, so a lot of the controls and complexities (and why it took so long to get to where we are) are because of that. Between being able to decide where your inference comes from and where the compute happens we had to grind through it all.
In terms of FedRamp compliance both AWS and Cloudflare have FedRamp options for inference, and we have the ability to do local inference as well which helps.
For the conversations themselves and FedRamp compatibility we do have BYOC at the larger enterprise levels.
To be honest though I think I'm most excited about the blend of local and cloud inference since it lets me offset the lighter work with local hardware.
@zackary_chapple thanks for the reply Zack!
sounds great. will be happy to dig into it more.
One of my favorite things about TAP is that you can build the random little tools you want without turning each one into a whole standalone product.
Miniapps just become part of the workspace alongside the people and agents already there.
I wanted to try that with something pretty far from the usual AI demo, so I adapted Mimaraka’s Daiza into a miniapp.
You drop in a transparent PNG, design an acrylic figure or keychain, preview it in 3D, and export the production files. It all runs locally inside TAP.
It was a fun one to build. Hopefully people bring their own weird, specific ideas too.
@nsttt need to show this off with some of the fabrication folks
I was actually blown away by how full featured and well built this platform is. I decided to try to make some things with their Mini App SDK, and I think "Mini App" might actually sell what it does short: its full on platform extension with robust integration hooks. I'm excited to experiment with using it to make this app mine, its an absolutely dank feature.
@ken_wheeler I need screenshots of the Slopometer you shared on the stream! That was pretty interesting
@ken_wheeler @zackary_chapple
Let's do round two of the podcast and see how you built the Slopometer!
I use The AI Platform to run simulations for students in my courses. The specialists and apps especially are helpful for me to give my students a near-real simulation of what it's like making product decisions. Great product!
@kentcdodds I can't wait to see the feedback from next week's cohort, there has been a ton of improvements made since the last cohort.
@kentcdodds I remember day one! It's great to have you on board!
Stand-ups assume everyone's awake at the same time, and once a team is spread across the Americas, Europe, and Asia that stops being true - someone's always attending at 6am, and the ones who can't make it get reconstructed secondhand. The usual workaround, "post your update async in a channel," lasts about a week before nobody's reading 15 people's updates and the ritual becomes theater.
It's gotten noisier in the agentic era, too. When part of your team is shipping through coding agents, one person's GitHub activity for a single day can be a dozen commits and three PRs - a raw digest of that isn't a standup, it's noise. So Pulse doesn't just dump activity: it drafts your update from it (or you type it fresh), and a specialist rolls the whole team's week into one summary, so a lead gets what actually mattered without wading through the firehose or hosting a meeting to extract it.
Worth calling out separately: we built Pulse as a miniapp on top of the platform, and that's the reason it moved as fast as it did. Roster, tasks, specialists, storage - Pulse doesn't own any of that infrastructure, it just asks the host for scoped grants to use it. That's the concrete answer to "what does building on a platform actually buy you": a full team-facing tool without writing auth, directory sync, or a task board from scratch.
@billwatson017 leaders in tech are going to have so much more visibility now.
My team considers me engineering adjacent, lol! I'm on the customer side and I hate CRMs. So I built my own with TAP. I can literally drop in context and it will route accordingly. It's going to be a game changer for me bc now I never have to worry about CRM hygiene. The CRM mini app will do it for me.
@zackary_chapple thank you for the assist!
You can point any clanker of choice at the Miniapp SDK and make something to share or for yourself. The SDK enables you to create pretty much anything on top of our platform, and makes it agent enabled.
I got really tired of having to switch apps to look at design files, so I decided to make a design file viewer. A quick miniapp to make reading and editing figma files possible inside of our TAP runtime. Was pretty fun to make and even more fun to see the agents go to town on it.
The AI platform gives you some amazing opportunities through mini apps. As a non-engineer, I was able to build a triage app that brings everything into one place and makes the whole triage process much easier.
Triaging tickets is always fun 😄 But it gets much better when you can keep the entire workflow in one platform: discuss tickets with all the stakeholders, turn those discussions into tasks that your team members or agents can pick up, bring their updates back in, and connect everything into one continuous flow.
For me, that’s one of the most exciting things about the mini app functionality — you can build the workflow you actually need without being an engineer.