Agent 37

Overview

Agent 37 is a managed cloud platform designed to host open-source AI agents — including OpenClaw, Hermes, and Claude Code — around the clock. It provides two complementary entry points: an operator-focused dashboard for one-click deployment and a Cloud API for developers who want to build agent-based products at scale. Underneath, both paths share the same managed infrastructure, eliminating the need to set up or maintain VPS servers, Docker containers, or complex orchestration layers. The platform is backed by Y Combinator and is positioned as a simple, reliable way to run autonomous agents without infrastructure overhead.

Each agent instance runs in an isolated, always-on sandbox equipped with a task board, web terminal, file browser, live desktop, and scheduled jobs — a complete environment for autonomous operation. Users can bring their own API keys or existing AI subscriptions, such as ChatGPT or Claude, and keys travel directly from the container to the model provider, preserving privacy and control. On higher plans (Plus and above), Agent 37 also bundles free models with usage limits on a rolling five-hour window plus a weekly cap. Connectivity is handled through Composio, enabling agents to interact with WhatsApp, Telegram, Slack, Discord, Gmail, GitHub, Google Calendar, and more than 1,000 other applications with one click.

For startups and product teams, the Agent 37 Cloud API simplifies the creation of per-user AI agents. A single POST /v1/instances request creates an isolated, always-on sandbox running Hermes, OpenClaw, Claude Code, or even a custom harness. Each end user gets their own disk and URL, and usage is billed per minute, starting from $1.99 per agent per month. The developer retains the API key and can white-label the entire experience under their own brand, making it suitable for embedding AI agents inside existing products or launching new agent-centric services.

Agent 37 places a strong emphasis on data privacy and customer control. It does not train models on any data the agent reads, writes, or says, and it does not collect conversations, files, or terminal sessions — a commitment that is written into its Terms. There is also no lock-in: all files and keys reside inside the container, can be downloaded at any time via the file browser or terminal, and cancellation is handled from the dashboard with access maintained for the paid period. Starting at $3.99 per month, Agent 37 serves both individual operators who need a personal always-on agent and engineering teams who want to scale agent fleets through a clean API.

Core capabilities

One-click Deployment

Deploy OpenClaw or Hermes instances directly from the dashboard without VPS or Docker setup.

Cloud API for Agent Products

Create isolated, always-on sandboxes for each end user via a single POST /v1/instances request; each sandbox has its own disk and URL.

Bring Your Own Keys/Subscriptions

Basic plan lets you use your own API keys that route directly from your container to the model provider, or sign in with existing ChatGPT/Claude subscriptions.

Bundled Models on Higher Plans

Plus and above include free bundled models with usage limits on a rolling 5-hour window and a weekly cap.

1,000+ Integrations via Composio

Connect WhatsApp, Telegram, Slack, Discord, Gmail, GitHub, Google Calendar, and more in one click.

Full Agent Workspace

Each instance includes a task board, web terminal, file browser, live desktop, and scheduled jobs.

Data Privacy Commitment

No training on user data and no collection of conversations, files, or terminal sessions, as stated in the Terms.

No Lock-in

Files and keys live inside your container and can be downloaded anytime; cancel from the dashboard and keep access for the paid period.

White-label Support

The Cloud API enables full white-labeling so you can run the agent experience under your own brand.

Per-minute Billing

API-created agents are billed per minute from $1.99 per agent per month, scaling cost with actual usage.

Use cases

Running a 24/7 personal AI assistant through OpenClaw or Hermes

Giving each SaaS user their own isolated AI agent via the Cloud API

Automating workflows with agents connected to Slack, Gmail, GitHub, or Google Calendar

Scheduling background jobs using the agent's live desktop and web terminal

Building and launching a white-label agent platform for clients

Prototyping Claude Code agents without local infrastructure

Scaling agent fleets with per-minute billing and no long-term commitments