Agent 37 Cloud

Product Introduction

Definition: Agent 37 Cloud is a managed cloud hosting platform and API infrastructure specifically designed for deploying, scaling, and persisting stateful AI agents like Hermes, OpenClaw, and ClaudeCode. It falls under the categories of AI-as-a-Service (AIaaS), agent hosting, and serverless container orchestration, providing a fully managed runtime environment instead of raw virtual machines.

Core Value Proposition: Agent 37 Cloud eliminates the operational complexity of running persistent agents by offering one-API-call provisioning of isolated, always-on agent instances for each end-user. It exists to let founders and developers ship vertical AI agent products to their customers without managing servers, writing adapter code, or handling infrastructure state, with costs starting from a transparent $3.44 per instance per month.

Main Features

1. Persistent Sandboxed Instances: Each API call creates a fully isolated, persistent environment for a single customer. This is technically implemented using gVisor for kernel-level container security and hard per-instance disk quotas. Unlike ephemeral serverless functions, the instance's files, memory, connected tools, and session state persist indefinitely until explicitly deleted, enabling truly conversational and continuous agent operations.
2. Unified REST API with Streaming: The platform wraps each agent's native CLI or SDK into a clean, consistent REST API. Developers can create instances (POST /v1/instances), send messages, and receive responses. All agent output—including text, reasoning steps, and tool activity—is streamed in real-time via Server-Sent Events (SSE), eliminating the need to build custom streaming adapters or WebSocket servers.
3. Preconfigured Agent Templates & Custom Docker Support: Users can instantly deploy industry-standard agents from a catalog (e.g., agent37-hermes for chat, browsing, and code; agent37-openclaw for channel integrations). For bespoke needs, the platform supports custom public Docker images, allowing developers to pre-install specific skills, models, and prompts. Each instance is automatically routable at a secure, HTTPS endpoint (https://{id}.agent37.app).
4. Integrated Model Gateway & BYOK Flexibility: The platform provides a built-in model gateway to access various LLMs, metering usage from a prepaid balance. Crucially, it also supports a Bring Your Own Key (BYOK) model, allowing developers to use their own API keys for models like Anthropic Claude or OpenAI GPT while still leveraging Agent 37's runtime and routing infrastructure, with per-instance spend caps for cost control.
5. Prepaid, Usage-Based Billing: Costs are calculated hourly based on allocated resources (vCPU, RAM, Disk) against a prepaid balance. There are no monthly fees or seat-based pricing; you pay only for the compute time your customer instances consume. Unused balance is prorated and returned hourly, aligning costs directly with usage and simplifying financial planning for SaaS businesses.

Problems Solved

1. Pain Point: The significant DevOps and infrastructure overhead associated with building a multi-tenant agent platform. This includes provisioning and managing fleets of VPSs or Kubernetes clusters, implementing complex state management for agent sessions, building custom APIs for agent interaction, and configuring networking, reverse proxies, SSL certificates, and persistent storage for every user instance.
2. Target Audience: AI Product Founders looking to ship SaaS products powered by autonomous agents; Vertical AI Application Developers building industry-specific tools (e.g., for legal, finance, or coding) who need to offer each client a dedicated agent workspace; and Independent Developers or small teams wanting to quickly prototype and launch agent-based products without a dedicated backend/DevOps engineer.
3. Use Cases:
   • Building a white-label AI assistant SaaS: Where each subscriber gets their own agent with memory and tool access.
   • Deploying a coding agent service: Offering each developer a cloud-based, persistent sandbox with Claude Code or a custom coding agent.
   • Creating specialized business automation tools: Using Hermes with custom plugins to provide clients with a managed agent that interacts with their specific SaaS tools (Gmail, Slack, etc.).
   • Rapidly prototyping agent-based applications: Validating a business idea by giving early access users instant, dedicated agent instances.

Unique Advantages

1. Differentiation vs. Traditional VPS/DIY: Unlike manually managing Mac minis, AWS EC2 instances, or Docker VPSs, Agent 37 Cloud is a turnkey solution. It completely abstracts away server patching, capacity planning, and glue code development. It provides a managed API, built-in networking/routing, and granular usage-based billing from day one, contrasting sharply with the fixed costs and high maintenance burden of self-hosted infrastructure.
2. Key Innovation: The core technical innovation is the "instance-as-customer" architecture delivered through a single API call. By combining a secure, persistent sandbox (gVisor), an auto-provisioned and routable endpoint, and an integrated model gateway into one atomic unit, it radically simplifies the multi-tenant agent deployment problem. This allows developers to shift their focus from infrastructure plumbing to product logic and user experience.

Frequently Asked Questions (FAQ)

How does Agent 37 Cloud handle data security and privacy for my customers' agent instances? Each customer instance runs in a kernel-level isolated sandbox powered by gVisor, ensuring strong isolation. Data, files, and memory are persistent per instance but completely separated from others. All communication occurs over HTTPS, and you control access via API keys. You can also opt for complete privacy by using your own model keys (BYOK), meaning agent interactions never touch Agent 37's model gateway.

Can I integrate Agent 37 Cloud with my existing application backend or frontend? Yes, absolutely. Agent 37 Cloud is designed for seamless integration. You make standard HTTP API calls from your backend (Node.js, Python, Go, etc.) to create instances and message them. Responses are streamed via SSE, which your frontend can consume directly for real-time updates. The platform's auto-generated instance URLs (https://{id}.agent37.app) can also be embedded within signed iframes for secure, direct browser access to terminal, desktop, or file browser views.

What is the difference between using Agent 37 Cloud and just running my agent on a basic VPS? A basic VPS gives you a server, but you must manually install runtimes, build and maintain an API layer, implement session persistence, configure networking/DNS/SSL, and set up monitoring and billing. Agent 37 Cloud provides all of this as a managed service. You get a ready-to-use API, automatic persistence, built-in security, global routing, and a pay-per-use pricing model in one package, saving significant development time and operational cost.