Introducing Micro Blocks: The Backend for Vibe Coding

Every time I've tried to build a small tool for myself, I've spent the first two days building the same boring things. A people table. A companies table. Some way to connect them. Then enrichment, then permissions, then a query layer.

By the time any of that works, I've lost interest in the thing I actually wanted to build.

Micro Blocks is the fix. It's the backend we built for Micro itself, and starting today it's available to developers.

What is Micro Blocks?

Micro Blocks is an agent-native platform for building personal software. You get typed objects like people, companies, and documents, typed property values, sharing settings, and a query language you don't have to write. People and companies come enriched out of the box, so there's no ingestion or data modeling step before you start building.

Our claim: it cuts the time and tokens required to build personal software by about 95%.

Email was already the super app

Here's the observation that started this.

Your inbox is a CRM. It's also shopping, travel, contracts, scheduling, hiring, and half your project management. All of that structure is already sitting in there. It's just trapped in a format that only renders one way, as a list of messages.

If you extract the objects out of email, you can render any interface you want on top of them. A pipeline. A travel dashboard. A reading list. A tracker for whatever weird thing your team cares about.

That's what we did to build Micro. The email client and the CRM you see are just interfaces over a shared object graph. That graph is Micro Blocks, and it's the same reason the CRM updates itself instead of asking you to.

We built the platform first and the product second. Now we're opening up the platform.

What you get on day one

When you start a project on Supabase or Firebase, you get a database. Powerful, and completely empty. Every concept you need has to be invented, modeled, migrated, and maintained by you.

Micro Blocks starts you further along:

• Typed objects. People, companies, documents, tasks, meetings, and more, already defined and already connected to each other.
• Typed property values. Properties know what they are, so your agent can inspect the model instead of guessing at column names.
• A relationship graph. The links between objects are first-class. You traverse them instead of writing joins across tables you built by hand.
• Enrichment. People and company data comes in enriched. You're not standing up an ingestion pipeline before your app does anything interesting.
• Sharing settings. Permissions are part of the platform, not a thing you bolt on at the end and hope you got right.
• A query language you didn't write. This one is underrated. Query languages are miserable to build and even more miserable to maintain.

The point isn't that these are impossible to build. It's that everyone builds them, every time, and they're never the interesting part.

Your agent gets memory, not just storage

This is the part I underestimated until we lived with it.

Most agent memory today is a pile of text. You dump transcripts and notes into a vector store, retrieve the chunks that look similar to the question, and hope the right facts came back. It works until the answer depends on a relationship rather than a phrase. Who introduced us. Which company she's at now. What we promised in the meeting before last.

Because Micro Blocks stores everything as objects in a graph, your agent can do both. Semantic search narrows the space, then the agent walks actual edges to get specifics. Brett works at Micro. Micro raised from Flybridge. Flybridge backed 40 other companies. One of those founders emailed you last week. That's four hops, and none of them require guessing.

Storage answers "what did I write down." A graph answers "how is this connected to everything else." The second one is what makes an agent feel like it remembers you.

We went deep on this in engineering a digital brain that dreams, which covers how we modeled the memory system on human memory architecture. The short version: your agent inherits that structure for free, on your data, the moment you build on Micro Blocks.

How Micro Blocks compares to Supabase, Firebase, and Airtable

	Micro Blocks	Supabase / Firebase	Airtable
Starting data model	Typed objects, ready to use	Empty tables	Empty bases
Relationships	First-class graph	Foreign keys you define	Linked records
People and company data	Enriched out of the box	Bring your own	Bring your own
Query layer	Included	You build it	Filters and formulas
Agent memory	Graph traversal plus semantic search	You build it	Not really
Permissions	Built in	You build it	Per-base sharing
Custom objects	Yes	Yes	Yes
Built for agents	Yes, including MCP	Not specifically	Not specifically

The honest summary: if you want a blank database with total control, use Postgres. If you want to ship a working tool this afternoon, start from objects that already mean something.

What you can build with it

The biggest proof point is Micro itself. The email client, the CRM, the tasks, the documents, the meeting notes, all of it runs on Micro Blocks. It isn't a side API we exposed for partners. It's the thing we use.

So you can build your own CRM shaped exactly the way you work, which is the app everyone rebuilds and nobody enjoys rebuilding. You can build a task manager. A client portal. An internal tool that only makes sense to your team. Personal software, for you or the people you work with.

If you've read why we built Micro, this is the same argument one level down. The product was our answer to tool sprawl. The platform is our answer to rebuilding the same foundation forever.

Is it really that flexible?

This is the first question every developer asks me, and it's the right one. An opinionated data model sounds great until it doesn't fit your problem.

You can create custom objects and custom properties, which means you can model basically any software you want. Our objects are a starting point, not a ceiling. And you can connect your own databases alongside them, so Micro Blocks doesn't have to own everything to be useful.

Where I'd push back on myself: if your app has nothing to do with people, companies, or work, you're not going to get much out of the graph. Micro Blocks is strongest when your software is about how people and organizations relate to each other. Most internal tools are.

Pricing and access

Micro Blocks is in beta. Access comes with a Micro account subscription, and you just have to stay inside the rate limits. No separate platform contract, no sales call.

Start building

Here's the fastest way to find out if this is for you, and it takes about ten minutes.

1. Sign up for Micro Blocks.
2. Drop docs.micro.so into your agent of choice.
3. Ask it to build something small. A tracker, a dashboard, one view you wish existed.

Don't start with the ambitious thing. Build the small tool you've been putting off, see how far your agent gets in one pass, and judge us on that.

More detail on objects, SDKs, and the API is on the Micro Blocks page.