7/ Under the hood it's PubNub, a decade of infra, 500M+ devices, 3T+ transactions/month. But all of those PubNub features live in the Blocks SDKs, CLI, and https://t.co/zaQRStv7hg. No PubNub knowledge necessary,. Free to start, and I'd genuinely love feedback: https://t.co/g9kuSFXx3S
We just launched https://t.co/fhF2qORuOj! It exists because of a problem that drove me nuts all year: thanks to Claude, building agents is easy now. Making an agent reachable by other people, reliably and securely? That part sucked. So @blocks_ai_ fixed it. ๐งต
6/ Scale: load balancing runs in the Blocks Network: we detect all available agent instances in sub-second time and fan tasks out across them. Want more throughput? Run another process. That's the whole playbook. Think a million agent instances on a million iPhones.
5/ Security is zero-trust all the way down: access tokens live ~60 seconds, refresh tokens are single-use, streaming credentials rotate per-task and per-stream. Revocation lands in seconds. And apps built on your agent inherit its rules: sharing an app can't leak agent access.
4/ From that one connection, your agent is instantly reachable by the people, apps, and agents you choose, whether it runs in a data center, on a laptop, or on an IoT device. Requests route over pub/sub channels, not IPs. Works with OpenClaw, LangChain, CrewAI, whatever you use.
3/ The core idea: your agent "dials out." It makes one outbound TLS connection on port 443 to the nearest point of presence, the same way a browser does. No listening port, no exposed service, no inbound anything. Nothing on your machine for an attacker to even find.
2/ Every agent we built hit the same wall: worked great on my laptop, then what? How can others access? Is it secure? What data can it touch? Is it even running right now? And no one wants to bet on a single agent framework/ecosystem ... things are moving too fast for that.
My self-hosted agent now has a real frontend I can hit from my phone.
But it's private by default. invite-only, free billing, nothing inbound exposed. Only the people I invite and I can access it, and my model and data stay on my machine.
no tunnel, no ports, no public URL to lock down.
https://t.co/brCP4e2ECN
@KayleyKiwi For us, we'd call an agent anything that receives a task and returns a result. It could be an LLM wrapper, an API integration, a data pipeline, a device controller, or custom business logic. Blocks doesn't care what's inside as long as it can receive work and produce output. ๐ค
const blocksAI = {
status: "loading...",
mode: "beta test",
access: "open now",
build: "in progress"
};
while (blocksAI.status !== "launched") {
console.log("Testing https://t.co/s307UzHFdT with early users...");
}
https://t.co/s307UzHFdT is loading. Beta test is underway.
Everyone's building agents; nobody's solved what happens after you build one.
Most agents won't reach the world. Yours will - with https://t.co/Dy9NwF8oAp.
Build your agent. It stays where it is. Blocks connects it to the world.
We're incredibly honored to announce that we have won Best-in-Show for Hospital Systems in the Transform Supply Chain & Resource Management category at the #HIMSS26 #EmergeExperience!
Todd Greene, our Co-Founder & CEO, will be speaking at the Emerge event on Wednesday, March 11th, between 11:50 AM - 12:35 PM. He'll be talking about the problems hospitals face and how PuNub can help solve them.
Learn more on our website: https://t.co/TMmpMx0eFw
Day 2 of #ECE25 is here! We've had many great conversations over the past few days and met many amazing people in the industry. Make sure to stop by our booth H60 to meet our awesome team at the event!
If you would like to learn more about PubNub (or need last-minute access to the event for a discount), be sure to visit our site: https://t.co/NiAdvI2iMb