Infrastructure for AI Agents

Give every agent its own Workspace

Powerful, lightweight, secure MicroVMs that boot in milliseconds. Drive them from the API, MCP, or the dashboard.

Works with your stack
OpenClawOpenClaw
CodexCodex
OpenCodeOpenCode
Claude CodeClaude Code
LangchainLangchain
n8nn8n
QdrantQdrant
NodeNode
PythonPython
GoGo
PostgresPostgres
ClickhouseClickhouse
OpenClawOpenClaw
CodexCodex
OpenCodeOpenCode
Claude CodeClaude Code
LangchainLangchain
n8nn8n
QdrantQdrant
NodeNode
PythonPython
GoGo
PostgresPostgres
ClickhouseClickhouse
oblien.workspace.create()ws-2b1cRunningAGENTData Analyst2 vCPU · 4 GB112msws-7f3aRunningAGENTCode Generator4 vCPU · 8 GB94msws-9d4eRunningAGENTQA Tester4 vCPU · 8 GB127msKernelStorageNetworkPER WORKSPACEFull RootIsolated Network per WorkspaceAuto TeardownAVG BOOT111msENCRYPTIONAES-256UPTIME99.9%
What is Oblien

What happens inside that workspace.

Spin one up for 30 seconds or run one for months. From the dashboard or the API.

01

Any agent, any framework

Claude Code, LangChain, CrewAI, AutoGen, or your own code - any image, running in seconds.

02

Docker images. Real isolation.

Your existing images work unchanged - each workspace gets its own hardware-isolated VM, not a shared runtime.

03

Dashboard or API

Click through the UI or drive everything with the SDK and REST API - same environment, no feature gap.

04

Spin up. Ship it. Drop it.

Thirty seconds then gone, or months as a live product with a custom domain - you decide.

How it works

From zero to running in seconds.

Three steps. Whether it runs for 30 seconds or 30 months.

01

Spin up an environment

Name it, pick an image, hit run. Seconds later it exists - isolated, networked, ready. No provisioning, no waiting.

02

Your agent takes over

It executes commands, writes files, runs workloads. Workspaces talk to each other over a secure internal network - agents, services, jobs, all wired together. Nothing exposed unless you say so.

03

Ship it or drop it

Expose a port, get an HTTPS URL, attach a domain. Or delete it - nothing left behind. Same decision, made in one click or one API call.

secure internal networkYour AgentSDK · REST API · DashboardWorkspace Aprivate · isolated · networkedWorkspace Bagent · service · jobWorkspace Cephemeral · or persistent

nothing exposed unless you say so

Use cases

Built for what you're building

Whether it's AI agents, code sandboxes, dev environments, or global workloads - one platform, zero compromise.

AI Agents

Give every agent its own machine

Concurrent agents get isolated workspaces with root access, browser, and network - they never interfere with each other.

  • Full OS with root access
  • Isolated from other agents
  • Auto-destroy on completion
RESULT
What you get

Everything an agent needs. Nothing it doesn't.

Each workspace is a self-contained machine — isolated, secure, and ready when your agent needs it.

Full Operating System

Every workspace is a real VM with its own kernel, filesystem, and init system. Not a container - a real machine.

Millisecond Boot

Workspaces are ready before your agent finishes its first inference. No cold-start penalty.

Hardware Isolation

Each VM runs on dedicated resources. A compromised workspace can never reach another.

Root Access

Install packages, compile code, run servers. Your agent has full sudo - no permission hoops.

AES-256 Encryption

Full-disk encryption at rest. Storage is securely erased on workspace deletion - no residual data.

Real Networking

Dedicated network interface per workspace. Outbound internet, DNS resolution, port exposure - not proxied.

Workspace-to-Workspace

Workspaces can communicate over an internal network. Build multi-agent systems that coordinate in real time.

Multi-Zone Routing

Route workspace traffic through the nearest zone for lowest latency. Switch regions instantly — no migration, no downtime.

Snapshots & Cloning

Freeze the entire machine state - memory, disk, processes. Restore later or clone a hundred times.

SSH Access

Direct SSH into any running workspace. Debug live, inspect logs, or hand off to a human operator.

Live Resize

Scale CPU, memory, and disk on a running workspace without restart. Pay only for what you use.

HTTPS Preview URLs

Expose any workspace port to the internet instantly. Auto-generated *.preview.oblien.com URL with TLS.

Ready-Made & Custom Images

Pick from Node.js, Python, Go, Rust, Redis, PostgreSQL and more — or bring your own image.

Managed Workloads

Run named background processes with independent lifecycles. Stream logs, monitor health, restart automatically.

Pause & Resume

Freeze the entire VM - processes, memory, everything. Resume instantly. Zero cost while paused.

TTL & Auto-Expire

Set a time-to-live on any workspace. Auto-stop or auto-delete when it expires. Extend with keep-alive pings.

Zero Trust Networking

All traffic blocked by default. Explicitly allow hosts, CIDRs, and ports per workspace. Dedicated outbound IPs available.

Real-time Metrics

Stream live CPU, memory, disk, and network stats via SSE. Track usage history across all workspaces.

Env Vars & Init Commands

Inject environment variables and startup commands at creation. Auto-delete on exit. Wait for full init.

Low Cost

Pay-as-you-go from $0.005/min. No idle charges on ephemeral workspaces. Free tier included.

See the API

A few lines to full control.

Create, control, and destroy workspaces with plain HTTP. No SDK required — any language, any framework.

AuthBearer token
Latency<50ms API response
FormatsJSON · SSE · WebSocket
workspace.ts
import { Oblien } from "oblien"
const client = new Oblien()
// Create a workspace
const ws = await client.workspaces.create({
image: "ubuntu-22.04",
vcpu: 2,
memory: 1024,
zone: "us-east-1"
})
// Execute a command
const res = await ws.exec("node -v")
console.log(res.stdout)
Internal API

End-to-end control from inside

A REST + WebSocket server runs inside every workspace on port 9990. Full programmatic control — no external APIs, no proxies.

Run commands

Execute synchronously or stream real-time output. Up to 50 concurrent tasks with stdin and configurable timeouts.

Filesystem access

Read, write, delete, and stream the entire filesystem. Recursive trees, gitignore filtering, and atomic writes.

Interactive terminals

Multiplexed WebSocket PTY sessions with binary I/O, resize support, 64 KiB scrollback, and auto-reconnection.

Full-text search

Ripgrep-powered regex search with context lines, file type filters, and case control. Results grouped by file.

File watcher

Watch directories for create, write, and delete events over WebSocket. 50 ms debounce with smart excludes.

Workspace transfer

Move files directly between workspaces over the internal network. Streaming multipart with integrity checks.

Your AgentSDK · REST · Dashboard:9990Internal API ServerREST + WebSocket · every workspaceExecuteRun commandsFilesRead · write · streamTerminalInteractive PTYSearchRipgrep full-textWatcherReal-time eventsTransferWorkspace ↔ workspacesecure · isolated · internal

REST, WebSocket, and SSH — same API, two access paths.

See the docs
Pricing

Transparent, simple pricing

No complicated pricing tables. Sign up free, get 500 credits, and launch your first workspace. A typical 1 GB workspace costs around $5/month.

1

Claim 500 free credits

Sign up and get 500 credits instantly. No card required.

2

Launch workspaces

Spin up VMs with the CPU, RAM, and disk you need.

3

Only pay while it runs

Per-second billing, no idle charges. Unlike traditional clouds, pay-as-you-go here is built to stay affordable — a full-time 1 GB workspace costs about $5/month.

500 free credits on signup

No credit card required

Start free
Monthly estimate
1 vCPU · 1 GB · 10 GB disk
Compute1 vCPU × 720 hrs
$3.60
Memory1 GB × 720 hrs
$0.72
Storage10 GB × 30 days
$0.50
Total~$5/mo
Per-second billing
No minimum commitment
Fair Billing

Your invoice will never explode

Costs cap automatically on every plan.

Runaway workloads, forgotten VMs, traffic spikes — none of it matters. Once you hit your monthly cap, everything keeps running at $0 extra.

Spending this monthSee all caps
$21.60FREE
$0$30 cap
Beyond your cap, workloads keep running — $0 extra

Per-workspace caps

Each workspace gets its own monthly spending ceiling

No overage charges

Usage beyond the cap is completely free

Monthly reset

Spending resets at the start of every billing cycle

Plans

A plan for every scale

Monthly credits, resource limits, and a cost cap included. Start with Hobby or scale to enterprise.

MonthlyAnnualSave ~17%

Hobby

$8/mo

For side projects and prototypes

Credits14.4k/yr
Workspaces5
Running pool4 vCPU · 4 GB
Max per WS2 vCPU · 2 GB
Disk10 GB

Pro

$24/mo

Ship production agents

Credits36k/yr
Workspaces10
Running pool8 vCPU · 8 GB
Max per WS4 vCPU · 4 GB
Disk10 GB

Scale

$124/mo

Run agents at team scale

Credits180k/yr
Workspaces50
Running pool64 vCPU · 64 GB
Max per WS16 vCPU · 32 GB
Disk50 GB

Enterprise

Custom

Custom infrastructure for organizations

CreditsCustom/yr
WorkspacesUnlimited
Running poolDedicated
Max per WSCustom
Disk

1 credit ≈ $0.01 · See full cost breakdown and calculator on the pricing page

Credit Packs

Top up anytime

Buy credits once — no subscription required. Larger packs give you more credits per dollar.

1k

credits

$10

Base rate

5k

credits

$45

Save 10%
Popular

10k

credits

$80

Save 20%

25k

credits

$175

Save 30%

100k

credits

$600

Save 40%

Or buy a custom amount from $10 to $10,000

Get in touch

Questions, partnerships, or feedback — we're here.

Email usDiscordTwitter

Ready to build?

Start with the free tier. No credit card. Your first workspace boots in under 130ms.

Read the docs

Stay in the loop

Product updates, engineering deep-dives, and new features. No spam.

Oblien - Where AI Agents Live & Operate | Workspaces, Tools, SDK