ARCHIL
Announcing Serverless SandboxesRead more →

AI infrastructure foragent platformsagent platforms

Run millions of stateful agents on top of your data with versioned disks, serverless sandboxes, and SQLite databases. Set up in minutes.

Start buildingRead the docs 
import { generateText, stepCountIs } from "ai";
import * as archil from "disk";
import { createDiskTools } from "@archildata/ai-sdk";
 
const { disk: agentDisk } = await archil.createDisk({ name: "agent-workspace" });
 
// Give the agent your whole filesystem as tools
await generateText({
  model: "anthropic/claude-sonnet-5",
  prompt: "Write a short story, format it nicely and save it as story.pdf",
  tools: createDiskTools(agentDisk),
  stopWhen: stepCountIs(30),
});
 
// Strongly consistent — read it back instantly, from anywhere
const story = await agentDisk.getObject("story.pdf");

Trusted by teamsshipping AI in production

DepotGammaGumloopCobberymicroAGIConcourse
Batteries included

Everything agents needto work with real data.

Sandboxes solve where agents run, but not how they get, keep, or share context. Archil does both.

Skills, customer data, and run logs converge through Archil into one filesystem mounted at /mnt/archil.CONTEXT SOURCESAGENT WORKSPACESkillss3://skillsREAD ONLYCustomer datas3://acmeREAD WRITERun logsdsk_runsREAD WRITECOMPOSECOMPOSED3 SOURCES · SHARED LIVE · 0 COPIES

Compose agentic context.

Context comes in layers — per-customer, per-agent, per-business — each with its own permissions. Archil combines them into one workspace that any number of agents can use at once.

Read the docs
research
support
analyst
research-agentready
Find Acme's renewal terms
Request dispatchedWaiting for filesystem
Agent reasoningParsing request
Access policyChecking inherited permissions
FilesystemOpening /customers/acme
In-place readReading contracts.pdf
completecontracts.pdf

Auto-renews annually

Notice period30 days0 bytes copied
/mnt/archilMounted · 0 copies
NameSourceSize
path/mnt/archil

Use data where it lives.

Make your S3 bucket or system of record available to agents as a native filesystem — without an ETL pipeline or a second copy to pay for, secure, and keep in sync.

Read the docs
agent-tool-run
preparing
Agent taskInspect Acme renewal terms
mounted /mnt/archil · context available
rg "renewal" customers/acme
3 matches · 84 ms
no command running · billing paused
python summarize.py matches.json
summary.json written · 412 ms
no command running · billing paused
node publish.mjs summary.json
artifact registered · 206 ms
sandbox released · filesystem preserved
2 vCPU4 GBactive compute
Billedmeter off

Run AI code on your data.

Give your agent serverless tools to manipulate their context with Bash, Python, or Node, without managing a separate sandbox.

Read the docs
agent-run / outputready
write /artifacts/q3-report.pdfwaiting for agent
northwind.ai/artifacts
S3 live
Northwind AI0/3 available
Generated artifactsS3 response
served via S3 APIno sandbox

Share what agents make.

Build experiences where your users collaborate with agents on the same files, through a strongly consistent S3 API.

Read the docs
agent-workspace · history
2 branches
  • run experimentreview/agent-7cp_a3f
  • roll back bad writecp_9d2
  • checkpoint before runmaincp_8f21
  • initial workspacecp_001
4 checkpoints2 branches
historyclean

Version agent data.

Agents make mistakes. Checkpoint the disk before they start, fork as many parallel attempts as you like, and keep the one that worked.

Read the docs
Data-intensive computing

Archilisbuiltforpetabyte-scaleproblems.

Run model training, genomics, and analytics through a unified SSD cache over your S3 data.

Benchmarks

Local-disk speed, S3 economics.

Archil keeps hot data on SSD and cold data in your bucket — so you only pay for what you use, not what you store.

Faster development work

measured by running git clone

100×

Archil
EFS / S3 Files1×

Faster storage operations

measured by performing 4 KB reads

30×

Archil
S31×
99.99%availability SLA
SOC 2 Type IIHIPAAEncrypted at rest & in transitData stays in your cloudBYOC available
Security details
User quotes

Builders are talking about Archil.

Mike Soylu
Founder
Adaptive
We've been testing Archil with our harness recently and it's hard to tell the difference in performance vs EBS — it's quite fast!
Nisola Shobayo
AI Engineer
Clay
Clay customers work with GBs or TBs of data. Getting that into a sandbox that spins up and dies in seconds used to mean slow staging. With Archil we just mount the filesystem and the data is there.
Garrison Snelling
Founder
ComputeSDK
Archil gave us a single filesystem layer that works identically across the 10+ providers ComputeSDK orchestrates. Our users stopped thinking about where their files live — that's exactly the point.
Andy Jefferson
Member of the Technical Staff
microAGI
They approached sandboxes from the perspective of data, not compute — and it's beautiful. This is serious engineering.
Darren Shepard
Chief Architect
Obot
This is my trifecta of a new-gen cloud — Archil for storage, compute alongside it, and a database. It just fits together.
Michael Chomsky
Founder
SetupClaw
Biggest surprise for me was Archil's speed. Incredible tech, and it should have more visibility with AI-native teams.
Mike Soylu
Founder
Adaptive
We've been testing Archil with our harness recently and it's hard to tell the difference in performance vs EBS — it's quite fast!
Nisola Shobayo
AI Engineer
Clay
Clay customers work with GBs or TBs of data. Getting that into a sandbox that spins up and dies in seconds used to mean slow staging. With Archil we just mount the filesystem and the data is there.
Garrison Snelling
Founder
ComputeSDK
Archil gave us a single filesystem layer that works identically across the 10+ providers ComputeSDK orchestrates. Our users stopped thinking about where their files live — that's exactly the point.
01 / 06