# Agent 37 Cloud (slim)

> Agent 37 Cloud hosts persistent Linux sandboxes for agent harnesses such as Hermes, OpenClaw, Claude Code and Codex, or a custom Docker image. A REST API creates instances, runs commands, manages files and sends messages to the agent inside.

- Full: https://www.anchorterminal.com/tools/agent37.md (~7,400 tokens) · this version ~1,930 tokens · JSON https://www.anchorterminal.com/tools/agent37.json · canonical https://www.anchorterminal.com/tools/agent37
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · Updated: 2026-10-08

**D · 46.1/100 · rank #650 of 722 · #10 in Code execution sandboxes · not agent-ready · confidence medium**


Assessment: Instances keep their disk until deleted, sleep when idle and wake on the next request, with compute published at $0.80 a vCPU-month and stable error codes on both APIs. The product launched on 29 September 2026. No status page, OpenAPI file, SDK library or published rate limits were found, and the starter credit needs a card.

## Facts

- Kind: HTTP API · vendor: Agent 37 Inc. · category: Code execution sandboxes · legal entity: Agent 37 Inc. · provenance 56/100
- Endpoint: `https://api.agent37.com/v1` (HTTP)
- Auth: API key · pricing: Pay per use · x402: no · licence: Proprietary service under Agent 37's terms of service. The template Dockerfiles and example apps on GitHub are MIT, and the `agent37` CLI on npm is Apache-2.0
- Probe metrics: not measured yet (probes haven't run)
- APIs: Hosting API at https://api.agent37.com/v1 (instances, templates and cloud builds, exec, budgets, backups, fork, signed URLs, public ports, crons, SSH keys, logs, metrics, usage). Agent API on each instance at https://{instanceId}.agent37.app (`/v1/responses`, sessions, files, models, health, version)
- Templates: Eight system templates (agent37-hermes, the default, agent37-openclaw, agent37-claude-code, agent37-codex, agent37-grok, agent37-opencode, agent37-pi, agent37-n8n), or a workspace template built from a Dockerfile or imported image of up to 8 GB
- Shapes: 2 vCPU and 4 GB, 4 and 8, 8 and 16, 16 and 32, with 2 to 80 GB of disk by shape. Default or Performance (dedicated cores). Burstable is experimental and by application
- Persistence: Disk kept until delete. Auto-sleep after 300 to 86,400 idle seconds (default 900) checkpoints the instance, and any request wakes it in a few seconds, or about two minutes with a fresh boot when it has to be rebuilt. Memory usually survives a wake but isn't guaranteed
- Backups: Nightly into seven rotating slots plus one on-demand slot, free, crash-consistent. Restore replaces data in place. Fork copies the disk into a new instance
- Exec: `POST /v1/instances/{id}/exec` runs through `sh -c` as the image user or root, waits up to 780 seconds, and caps stdout and stderr at 512 KB each with a `truncated` flag
- Browser: agent37-hermes and agent37-openclaw ship a headless Chromium the agent drives. A desktop template with noVNC on port 6901 is a separate cloud build from the examples repository
- Isolation: The vendor says each sandbox runs under gVisor with its own filesystem and disk quota, that egress reaches only the public internet, and that hosts have no public hostname (https://www.agent37.com/cloud)
- Limits: One instance before the first top-up, 10 after it, 50 at $100 of total top-ups and 200 at $250. Three concurrent template builds. One on-demand backup per instance every 15 minutes. No request rate limits published
- Errors: `{error: {code, message}}` on the Hosting API, with `param`, `hint` and `response_id` added on the Agent API, and flat `{error: "code"}` strings for transport failures. 18 Hosting API codes, 13 transport codes and 16 Agent API codes are listed
- Tooling: `agent37` CLI on npm, version 0.4.0 of 11 September 2026, Apache-2.0. Docs examples in curl, Python requests and JavaScript fetch. No SDK library or OpenAPI file found
- Data: Instances run on dedicated servers in the United States. A deleted instance's data stays in backup storage for seven days. Request logs are kept up to 30 days. The privacy policy names 12 kinds of third-party service
- Certifications: The vendor says a SOC 2 Type I report is available on request under NDA and that Type II is in progress. The trust centre refused our reader
- Prices: vCPU $0.0011 per vCPU-hour; Default instance (2 vCPU, 4 GB, 4 GB disk) $0.0065 per session-hour; Disk $0.09 per GB per month

- Scores: Reliability 27, Performance pending, Schema & documentation 57, Agent ergonomics 60, Security & auth 49, Payments & pricing 20, Task success pending, Maintenance & community 65, Transparency & trust 56 · total over the 7 assessed categories
- Why: Reliability, Graded as a hosted service. · Schema & documentation, No OpenAPI or other machine-readable contract found. · Agent ergonomics, Exec output is capped at 512 KB a stream with a `truncated` flag, logs take a `tail` and chat can return one JSON body instead of a stream… · Security & auth, One workspace-wide `sk_live_` key, revocable, stored as a one-way hash per the privacy policy, with an optional IP allowlist of up to 20 add… · Payments & pricing, No x402, MPP or L402 on the API (0). · Maintenance & community, Scored on the closed-service scale. · Transparency & trust, Closed service with dated terms (4 August 2026).
- Sources: 23, open questions: 7, both in the full twin
- Capabilities: sandbox.persist, sandbox.code, sandbox.fs, sandbox.browser
- JSON: https://www.anchorterminal.com/api/v1/tools/agent37.json
- Verify (for the vendor): the badge `https://www.anchorterminal.com/badges/agent37.svg` or a link to https://www.anchorterminal.com/tools/agent37 from a page on agent37.com or one of its subdomains, or the README of github.com/agent37-platform/templates, then `POST https://www.anchorterminal.com/api/v1/verify` `{"slug", "url"}` or `verify_listing` at /mcp; re-checked weekly, no effect on the grade. Snippets in the full twin.

## Before you call it

1. Send the key as `Authorization: Bearer` to api.agent37.com and as `X-Agent37-Key` to the instance URL. The Bearer form on instance URLs is deprecated
2. Poll `GET /v1/health` on the instance URL for `"healthy": true` before the first message. `status: running` only means the machine is up
3. Check whether `error` is a string before reading `error.code`. Transport errors are flat strings, and a failed turn returns 200 with `status: "failed"`
4. Don't blindly retry an exec that timed out after 13 minutes. The command keeps running, so a retry starts a second copy
5. Set `auto_sleep: true`, keep anything that must survive a sleep in a file, and allow a first-byte timeout of three minutes for wakes

## Connect

```bash
curl -X POST https://api.agent37.com/v1/instances \
  -H "Authorization: Bearer sk_live_..." \
  -H "Content-Type: application/json" \
  -d '{ "budget": { "credit_micros": 1000000 } }'
```
Full config and headless snippets are in the full page. Through letme (picks today, calling later): https://letme.dev/agent37

## Similar tools

| Tool | Grade | Score | Shared capabilities | Slim |
| --- | --- | --- | --- | --- |
| E2B | B | 68.3 | sandbox.code, sandbox.fs, sandbox.persist, sandbox.browser | https://www.anchorterminal.com/tools/e2b.min.md |
| Daytona | B | 64.3 | sandbox.code, sandbox.fs, sandbox.persist, sandbox.browser | https://www.anchorterminal.com/tools/daytona.min.md |
| Microsoft Execution Containers | BB | 76.3 | sandbox.code, sandbox.fs, sandbox.persist | https://www.anchorterminal.com/tools/microsoft-execution-containers.min.md |
| Modal Sandboxes | BB | 75.5 | sandbox.code, sandbox.fs, sandbox.persist | https://www.anchorterminal.com/tools/modal-sandboxes.min.md |
| Vercel Sandbox | B | 69.6 | sandbox.code, sandbox.fs, sandbox.persist | https://www.anchorterminal.com/tools/vercel-sandbox.min.md |


## Panel reviews (0, desk reviews from public material, no calls made)