Agent 37 Cloud

Agent 37: The Ultimate Infrastructure for Persistent AI Agent Sandboxes

In the rapidly evolving landscape of artificial intelligence, developers face significant challenges when deploying stateful, customer-facing agents. Agent 37 emerges as a comprehensive solution, providing persistent sandboxes for Hermes, OpenClaw, and any other custom agent architecture. By using Agent 37, developers can bypass the complexity of managing infrastructure and focus entirely on building high-value agentic experiences.

What is Agent 37?

Agent 37 is a cloud-based platform designed to host always-on AI agent instances. Unlike traditional hosting environments that require manual setup of virtual private servers (VPS), Agent 37 offers a managed infrastructure where every customer receives their own dedicated environment. This approach ensures that files, memory, and connected tools remain persistent across multiple sessions until the instance is explicitly deleted.

Agent 37 Cloud is built to scale, offering a streamlined architecture that allows you to provision an instance for a customer with a single API call. Whether you are running a standard model or a complex custom agent, Agent 37 provides the stability and persistence required for production-ready applications.

Key Features of Agent 37

Agent 37 is packed with features that simplify the deployment and management of AI agents. Below are the core capabilities that make the platform a leader in the agentic infrastructure space:

Persistent Architecture

With Agent 37, every customer gets their own agent instance. This architecture ensures that a single POST request provisions a unique environment where the agent's files, memory, and connected tools stay put. This persistence is vital for long-running tasks and continuous customer interactions.

Kernel-Level Isolation

Security is a top priority for Agent 37. Every instance runs under the gVisor runtime, providing hard kernel-level isolation. Additionally, each Agent 37 instance is governed by a strict per-instance disk quota, ensuring that your resources are protected and your costs are controlled.

Zero-Configuration Networking

Agent 37 eliminates the need for manual proxy wiring. Every port of every instance is automatically routable via a secure URL (https://{id}.agent37.app). This includes:

• No reverse proxy to manage.
• No DNS configuration required.
• No SSL certificates to maintain.
• Access behind your own key or a signed URL.

Advanced Agent Capabilities

Every Agent 37 instance acts as a fully functional computer for your agent, supporting:

• Any Model: Connect to any AI model through the Agent 37 gateway or bring your own model keys.
• 1000+ Integrations: Connect seamlessly with Gmail, Slack, Notion, and more, all routed and metered for you.
• Custom Skills: Preinstall prompts, tools, and skills via custom templates.
• File Management: Upload and download files directly from your backend to the Agent 37 instance.
• Execution: Run shell commands and scheduled cron jobs inside the sandbox.
• Live Monitoring: Access streaming events for text, reasoning, and tool activity via Server-Sent Events (SSE).

How to Use Agent 37

Getting started with Agent 37 is designed to be developer-friendly. You can put a working agent in front of your first customer today by following these simple steps:

1. Create an Instance

You can provision a new Agent 37 instance using a simple cURL command, or via the Python and TypeScript SDKs. For example, to create a Hermes-based instance:

curl https://api.agent37.com/v1/instances \
  -H "Authorization: Bearer sk_live_..." \
  -d '{
    "template": "agent37-hermes",
    "resources": { "cpu": 2, "memory": 4, "disk": 6 }
  }'

2. Message the Instance

Once the instance is running, you can immediately begin sending messages and receiving responses. The Agent 37 Cloud handles all the routing and state management internally.

3. Embed and Monitor

You can embed a live terminal, desktop, or file browser using signed URLs, allowing your customers to interact with their Agent 37 instance in real-time.

Available Agent 37 Templates

Agent 37 provides several templates to jumpstart your development, or you can build something entirely unique.

• Hermes: The default template. It supports chat, browsing, code execution, and file management. It is ready to message the moment it is created.
• OpenClaw: A specialized template for channels-and-models agents, hosted behind the same consistent Agent 37 API.
• Claude Code (Coming Soon): Anthropic’s coding agent on its own dedicated computer.
• Custom Templates: You can use any public Docker image and register it under a custom name, building on the Hermes base or starting from scratch.

Agent 37 Use Cases

Agent 37 is built for production environments where reliability and branding are essential.

• White-Label Agent Products: Since you hold the key, your customers see only your brand, never Agent 37. It is white-label by default.
• SaaS Integrations: Use Agent 37 to connect your app to over 1,000 external tools like Slack and Notion for your customers.
• Automated Workflows: Leverage scheduled runs (cron) to give an agent a schedule; it will run unattended while you read the results from your backend.
• Developer Sandboxes: Provide customers with a persistent computer environment where they can run code, manage files, and interact with tools across multiple sessions.

Pricing and Managed Hosting

Agent 37 offers a transparent "Pay for compute, not seats" pricing model. This is significantly more cost-effective than a basic VPS and includes the agent, API, and routing.

Compute Costs

• vCPU: $0.80 / vCPU / month
• RAM: $0.70 / GB / month
• Disk: $0.09 / GB / month

For a standard configuration (1 vCPU, 3 GB RAM, 6 GB Disk), the cost is approximately $3.44/month or $0.0047/hour.

Flexible Billing

• Prepaid Balance: Instances are billed hourly from a prepaid balance.
• Prorated Refunds: If you delete an instance, unused prepaid time is returned to your balance, prorated to the hour.
• Free Tier: Your first Agent 37 instance is free, and no credit card is required to start.

Frequently Asked Questions (FAQ)

Does Agent 37 require me to use their model keys? No. You can bring your own keys or choose to meter models from your Agent 37 balance with per-instance spend caps.

How is security handled for each customer? Agent 37 uses kernel-level isolation through gVisor to ensure that each instance is completely separate and secure from others.

Can I expand the storage of my instances? Yes, you can add disk space to any Agent 37 instance for $0.09 per GB per month.

Is there a way to limit costs per customer? Absolutely. You can top up your balance and set a specific spend cap for every Agent 37 instance to prevent unexpected overages.

How do I get help setting up my first agent? You can start for free immediately, or you can book a 30-minute session to talk to the founder and get set up live.