Archil

The file system that connects AI to its data. We power the agents, platforms, and models you love.

[Archil @archildata]

See more information in our documentation here: https://t.co/DS2BU6cVPx

[Archil @archildata]

Today, we're extremely excited to announce Checkpoints and Branching functionality for Archil disks.

Checkpoints allow you to create points-in-time on your disk that you can rollback to, or create new branches from -- all in milliseconds.

This makes it possible to create branching versions of any software that runs on Linux: large data sets, SQLite, or even Postgres, and it's critical functionality for people building coding agents that need to explore different outcomes in parallel, or rollback bad changes.

Checkpoints and branching is available today in all AWS regions.

[Tigris Data @TigrisData]

Agents need safe ways to try, fail, fork, and iterate.

On June 9, we’re co-hosting Fork it, Branch it, Bop it with @PlanetScale in SF — a demo night on forkable infra for agent builders.

Featuring demos from @archildata, @agentuity, @daytonaio, and @mesa_dot_dev.

https://t.co/EjFX5orkQE

[Eugene @GenexRoberts]

actively recruiting for Founding GTM @archildata

we’re building the cloud filesystem for AI — persistent storage + compute infrastructure for agents, training, and AI-native applications.

if you’ve sold developer tools, cloud infra, AI/ML platforms, or technical enterprise products — and want to help build a category-defining company in AI infrastructure, let’s talk.

[Archil @archildata]

Today, we're launching the ability to view and run SQL against SQLite databases directly from the Archil console.

Developers on the forefront of agent development have told us that their agents need a mix of unstructured (Markdown) and structured (database) files in order to deliver the highest performance.

With this launch, Archil makes it simple for developers to debug and inspect their agents state, even if some of that state is stored in database files.

Available now on the Archil console.

[Archil @archildata]

See more in our documentation here: https://t.co/KjwmsGoSnJ

[Archil @archildata]

Today, we're announcing multi-disk Serverless Execution. This allows you to build agents that have access to a bash tool (with Python, Node, and more) that can compose data from different file systems, in a directory structure that you specify.

This simplifies how teams building enterprise agents manage data when, for example, a central team is building and updating skills independently from file systems that are for customer data.

[Archil @archildata]

See more information in our docs here: https://t.co/yY9Y22xGp1

[Archil @archildata]

Introducing the Archil S3 API. This API makes it simple for users to get fully-consistent access to their data from both a file system (for agents) and an S3 API (for existing applications or users).

With this launch, you can now easily view Archil files in the console and share them between coworkers or hand them off to Claude, even while they're being generated in the cloud.

The Archil S3 API is available to all Archil file systems today.

[Dalton Caldwell @daltonc]

As part of leading the Series A for @archildata, I sat down with the founder/CEO @jhleath

We discuss:
- What exactly Archil is, and why anyone building agents should evaluate it
- Why LLMs understand filesystems better than pretty much any other technology
- Hunter's 10 year career at AWS working on filesystems
- A case study of how Clay is using Archil
- What it was like raising a Series A from Standard Capital

and more!

[Archil @archildata]

@Standard_Cap See more in our announcement post here: https://t.co/W3l7WU4nLe

[Archil @archildata]

Today, we're announcing our $11M Series A raise, led by @Standard_Cap in order to build the layer that connects AI to its data.

The next generation of agentic applications are inherently stateful in a way that existing primitives don't solve for. Archil solves this for agents by giving them an infinite, high-performance file system that they can use to directly run bash and Linux programs.

[Archil @archildata]

For more details on the disk package, please see the npm documentation here: https://t.co/gX5tGvVjmv

[Archil @archildata]

Today, we're excited to announce the "disk" npm package as a way to manage your Archil file systems. Developers and agents can now use the disk package and CLI to create, manage, and execute code directly on their Archil disks. 🧵

[Archil @archildata]

For example, you can now create disks with "npx disk create" and immediately execute code on the disk using "npx disk execute".

We think this will massively simplify how users work with Archil without needing to first do complex configuration.

[Hunter Leath @jhleath]

it turns out that when we added the ability to execute code directly on an Archil disk, we inadvertently built the fastest service to spin up sandboxes for agents that use a real VM.

last week, it was exciting to see @rauchg and @cramforce celebrate being in this position. i was surprised when our exec endpoint not only outperformed the Vercel sandboxes, but was faster by over 2x.

the ability to spin up compute quickly is obviously becoming fundamental as agents need the ability to run code in the cloud. however, Archil is not a sandbox company, and there's a real difference in our offering.

unlike other providers, Archil runs user code *on an individual disk*. this means that every one of these VMs that spun up had concurrent access to an infinite file system (synced to S3), and start with the same low latency regardless of the amount of data in that file system.

we think this is going to be a huge unlock as agents move from single-player to multi-player and work on larger and larger amounts of context.

finally, a note to our customers.

we want to be clear that this isn't a celebration for us. i consider myself massively *disappointed* in the numbers that we put up today. celebrating 150ms to spin up a VM further fuels the discourse that there is some kind of "compute ladder" and that the vast majority of software will be rewritten to typescript(?) or compiled to WASM(?).

we think that this is fundamentally the wrong direction for the industry. instead, we think that it's better to do hard thing without compromises.

we will not stop until you can use `exec()` on an Archil disk to get a full Linux environment in single-digit milliseconds. this is what the next generation of agentic applications need, and this is what we intend to deliver.

[Archil @archildata]

BREAKING: Archil is now the fastest way to run code in the cloud for agents with a real VM security boundary, as tested by @computesdk.

Simply set up a disk, and run ".exec()" in order to execute code on it.

[ComputeSDK @computesdk]

🚨🚨🚨

Benchmark Friday is back and @daytonaio is STILL at the top.

However, we have some new competition for second place w/ @archildata dismantling @vercel this week by entering their serverless disks in the benchmark.

[Archil @archildata]

Introducing serverless execution. You can now give agents a bash tool that supports running bash directly on its context, without needing to manage a separate lifecycle for sandboxes.

Generally available now to all file systems in AWS regions.

https://t.co/Rr3XNE9MsZ

[Hunter Leath @jhleath]

https://t.co/9uKSZvVW0L

[Archil @archildata]

Get access to the Mac application today using the install instructions available here: https://t.co/nz4nm71qy9

[Archil @archildata]

Today, we're excited to release Archil's native, local application for Macs.
You can now, for the first time, get access to the same files as your cloud applications and sandboxes are using, so that you can locally debug, edit, and inspect what's going on with tools like Claude.