Agent 37

Persistent sandboxes and Linux VMs for AI agents, code, and services at 95% lower cost than competitors.

Overview

Agent 37 is a cloud platform that provides isolated sandboxes and full Linux VMs for hosting AI agents, running code, and deploying services. Each customer gets their own persistent sandbox environment where agents can execute tasks, access files, run Docker containers, and connect to external tools—all at significantly lower cost than alternative sandbox providers.

The platform is designed specifically for agent applications, supporting popular agent frameworks like Hermes, OpenClaw, Claude Code, and custom Docker images. Sandboxes persist across sessions, allowing agents to maintain state and files over time. Compute is metered by the minute while running, and disk storage is retained separately, with automatic sleep/wake functionality to minimize idle costs.

Agent 37 includes built-in security with kernel-level isolation via gVisor, encrypted data at rest and in transit, and SOC 2 Type I compliance. The platform also offers a white-label dashboard for building multi-tenant agent products, complete with chat interfaces, file management, messaging app integrations, and billing controls.

Key features

Persistent Linux sandboxes and full VMs
Per-minute compute billing, disk retained separately
Automatic sleep/wake on idle
gVisor kernel-level isolation
Docker and Docker Compose support
Android 14 emulator with ADB
HTTPS port access and signed preview URLs
Multi-tenant white-label dashboard
Messaging app integrations (Telegram, WhatsApp, Slack, Discord)
1000+ app connections via Composio

Pros

95% cheaper than most sandbox competitors at published rates
Persistent storage across sessions and restarts
True Linux VMs with root access and full software installation
Flexible billing: pay only for compute while running, disk while retained
Built-in multi-tenant dashboard for white-label products
Comprehensive security with kernel isolation and encryption
Supports multiple agent frameworks and custom Docker images

Cons

Pricing comparison assumes specific workload patterns; actual costs vary by usage
Requires API key management for programmatic access
White-label dashboard requires Supabase setup and deployment
Limited to Linux-based workloads (though Android emulation available)

Use this if

You need to host AI agents with persistent state, run isolated code workloads at low cost, build multi-tenant agent products, or require Docker and Linux VM capabilities with flexible per-minute billing.

Skip this if

You need Windows environments, managed Kubernetes, or serverless functions without persistent storage. If you require monthly subscription pricing or included model tokens, consider alternatives.

Best for

Hosting AI agents with persistent stateRunning code and services in isolated environmentsBuilding multi-tenant agent productsCost-sensitive sandbox workloadsDocker and containerized applicationsAndroid automation and testing

Alternatives

E2BDaytonaModalFly.ioAWS EC2DigitalOcean

Compare Agent 37 alternatives

At a glance

Free plan: Yes
Trial: Yes
Paid plans: Yes
Setup time: Minutes (API-based provisioning)
Learning curve: Medium

Get started

curl https://api.agent37.com/v1/instances -H "Authorization: Bearer sk_live_..." -H "Content-Type: application/json" -d '{"template":"agent37-hermes"}'

Standard 2 vCPU / 4 GB RAM / 4 GB disk: $0.80/vCPU-hour, $0.70/GB RAM-hour, $0.09/GB disk-hour, metered per minute. Estimated $4.76/month for 730 running hours. Performance tier costs 4x compute. New accounts receive $5 starter credit after adding a card (no charge until top-up). No subscription fees, plan minimums, or seat charges. Model tokens and integrations meter separately from the same balance. Source

Curated by Newtools

Added and checked by the editorial team

Sources