Skip to content
TrustList
A3
Application Development Software

Agent 37

New· 4

Cloud sandboxes for AI agents, billed by the minute at low unit rates, for running code, agent harnesses, browsers and desktops, with a separate sandbox for each customer of the product built on it.

  • Artificial Intelligence - AI
  • DevOps
  • Docker

TrustList summary

Not yet independently verified. Dated from the Y Combinator launch post of 11 November 2026, as the site gives no launch date. The cost comparisons and benchmark ranks are the maker's own, and the pages name no city or country. We will update this when it can be confirmed, and remove this note.

Agent 37 rents out the isolated computers that AI agents need to run code, browse and use a desktop, and competes almost entirely on price. Its launch post argues that sandboxes have "no moat" and promises to refund the difference to anyone who finds a cheaper one.

Per-minute rates

Compute is metered by the minute while a sandbox runs, and disk while it is kept:

  • vCPU: $0.0011 per vCPU-hour
  • Memory: $0.00096 per GB-hour
  • Disk: $0.00012 per GB-hour

The launch post works this through for a sandbox with 2 vCPUs and 4 GB of memory left running for a month, at $4.76. New accounts get $5 of credit, and startups can apply for up to $10,000 in credits.

What runs inside

A sandbox can host a ready-made agent harness, among them Hermes, OpenClaw, Claude Code, Codex and OpenCode, or a team's own, as well as background jobs, Docker containers, and desktop and browser sessions. The pitch to software companies is isolation per end customer: each customer's tools, files and credentials sit in a separate sandbox. Apps inside a sandbox can be exposed over HTTPS on a provided address with a required key, or on the customer's own domain. The site says the service powers agents for more than 250 startups.

Key facts

Maker: Agent 37 Launched: Y Combinator launch post, 11 November 2026 Billing: prepaid, metered by the minute Pricing: $0.0011 per vCPU-hour, $0.00096 per GB-hour of memory (site, read 12 November 2026) Backed by: Y Combinator

Categories & features

  • Artificial Intelligence - AI
  • DevOps
  • Docker

Trust Score

4/ 100
Trust Score: New

An earned signal from verification, reviews, awards, transparency and engagement - the vendor can't buy it.

Verification
0/100 · 20%
Reviews
0/100 · 30%
Awards
0/100 · 15%
Transparency
18/100 · 20%
Engagement
0/100 · 15%
Joining soon
Recommendations
Coming soon
Complaints
Coming soon

Updated 11/12/2026

Request a demo or quote from Agent 37

Protected by reCAPTCHA - Google Privacy Policy and Terms apply.

By sending, you agree we may share your request and contact details with the provider once you confirm your email.

Reviews

Write the first review of Agent 37

Used it? Your experience helps other buyers decide.

Write a review

Questions & answers

No questions yet. Be the first to ask about Agent 37.

Agent 37 alternatives

Similar listings buyers compare against this one.

MIT-licensed runtime of single-request serverless actors, each with its own SQLite database and WebSocket server, written in TypeScript on a Rust runtime; run on Terse Cloud or self-hosted. Billed by the second for CPU and memory requested.

An AI coding agent for C, C++ and CUDA that works from the real build and call graph, takes bugs from Slack, Jira or the command line, and returns a pull request with build, test, sanitizer and performance results attached.

Analytics tool showing how people use Model Context Protocol servers after installing them

A Bolt.new agent that runs on open-source models and gives more usage in return for training data

Packs a whole application and its data into one portable file that opens without a server or cloud account

MIT-licensed C library for terminal emulation, taken from the core of the Ghostty terminal. Version 1.0 declares the library and its C ABI stable, with no copyleft code in its dependencies.

Routes each coding agent request to the cheapest model likely to get it right across a team's existing subscriptions

API that turns a URL into Markdown written by a vision model, with the page's hidden JSON and a citable snapshot. Priced at $1 per 1,000 fetches per its launch post and pricing section.