LelantosLelantos

Introduction

Lelantos is a secure sandbox platform for AI agents with sub-90ms boot, EU data residency, and E2B compatibility.

What is Lelantos?

Lelantos is a secure, high-performance sandbox platform purpose-built for AI agents. It provides on-demand isolated compute environments powered by Firecracker microVMs running on Hetzner bare metal servers in Germany -- giving you EU data residency by default.

Lelantos is fully compatible with the E2B SDK, so you can migrate existing E2B workloads with minimal code changes while gaining EU hosting, longer session lifetimes, and lower costs.

Key Features

EU Data Residency

All sandboxes run on Hetzner bare metal in Germany. Your data never leaves the EU.

Firecracker Isolation

Every sandbox runs in its own Firecracker microVM with dedicated kernel, providing hardware-level security isolation.

Sub-90ms Boot

Warm pool of 30 pre-warmed sandboxes enables sub-10ms assignment. Cold starts boot in under 90ms.

E2B Compatible

Drop-in replacement for E2B. Use the same SDK patterns and migrate with a single import change.

Per-Second Billing

Pay only for what you use. vCPU at 0.98 milli-cents/sec, RAM at 0.315 milli-cents/sec per GiB. $50 free signup bonus.

24-Hour Sessions

Default sandbox lifetime of 24 hours -- 48x longer than E2B's default. Extend or set unlimited with a single API call.

Core Concepts

Sandbox

A sandbox is an isolated compute environment backed by a Firecracker microVM. Each sandbox has its own CPU, memory, disk, and network stack. You can execute commands, upload/download files, and access running services via a unique URL.

Default resources per sandbox:

  • 2 vCPU / 512 MB RAM / 20 GB disk
  • Full internet access with pre-configured DNS
  • Unique proxy URL: {port}-{sandboxID}.lelantos.ai

Template

A template defines the base environment for a sandbox. Templates are built from Dockerfiles and produce a root filesystem image. Lelantos provides six pre-built templates (base, python, node, bun, playwright, code-interpreter) and supports custom templates with configurable CPU (1-8 vCPU) and memory (128-8192 MB).

Architecture Overview

Client (SDK / cURL)
       |
       v
  API Server (api.lelantos.ai)
       |
       v
  NATS Message Bus
       |
       v
  Worker Node (bare metal, Hetzner DE)
       |
       v
  Firecracker microVM
  (isolated sandbox)
  1. Your application creates a sandbox via the REST API or SDK.
  2. The API server dispatches the request through NATS to an available worker node.
  3. The worker boots a Firecracker microVM from the template's root filesystem image.
  4. The sandbox is assigned from the warm pool (sub-10ms) or cold-started (sub-90ms).
  5. You interact with the sandbox via commands, file operations, and HTTP proxy.

Lelantos vs E2B

FeatureLelantosE2B
Data residencyEU (Germany)US
Default session timeout24 hours5 minutes
Signup bonus$50 (no credit card)--
Per-second billingYesYes
Self-hostableYes (planned)No
vCPU pricing0.98 milli-cents/secHigher
Custom templatesYesYes
SDK compatibilityE2B SDK compatible--

Next Steps

On this page