# Tencent Cloud CubeSandbox

Official pricing: https://github.com/TencentCloud/CubeSandbox  
Category: agent-sandbox · Isolation: vm

## Pricing regimes (raw)

- **On-demand** (resource)
- **Self-hosted CubeSandbox (Apache-2.0)** (resource) [alt]

## Features

Yes: snapshots, memory snapshots, fork/clone, pause/resume, persistent disk, volumes, idle auto-stop, ≥24 h sessions, custom image (Docker or snapshot), start from your own snapshot, your own Docker/OCI image, full VM (own kernel), Docker inside, browser, desktop GUI, preinstalled agents, HTTPS preview URLs, egress allowlist, open internet, static egress IP, self-hosting, open source, arm64, Python, Node.js, credential injection, HTTP method/path egress rules, wake on request, live fork (no pause), memory fork, EU data residency, automatic snapshots, snapshots on demand, hosted agent API (their own agent), gVisor or VM (no shared kernel), inbound access rules, firewall inside (nftables), secret proxy, secret proxy for any API, extra volumes, shared volumes

No: nested virtualization, computer-use API, browser + desktop control, anti-bot stealth, CAPTCHA solving, residential IPs, SSH, public IPv4, BYOC, SOC 2, HIPAA, SSO, GPU, Windows, macOS, live resize, MCP server, agent harness API

Unknown: everything else. Evidence (source + quote) per feature: https://battleships.dev/data/providers/tencent-cubesandbox.json → feature_evidence

## Caveats

- Official SDK/public Tencent Agent Sandbox product. Chinese-cloud pricing and shape not reliably extracted in this pass; unknown is not zero.
- Not fully priceable in ENGINE_CARD: unknown USD resource rates or machine shape. Null is not zero; do not rank as a free provider.
- Features set to null were not verified. No sandbox was purchased or tested; feature/performance statements are documentation claims. Null maximums are unknown unless a cited note explicitly states unlimited.
- A complete monthly bill requires known snapshot/egress/plan limits; do not silently treat missing ancillary charges as free.
- Re-checked 2026-09-28: the Agent Runtime product page (https://cloud.tencent.com/product/ags) and its docs (https://cloud.tencent.com/document/product/1814: overview, quick start, quotas, FAQ) publish no per-resource rate or billing page; the page only says billing is on demand ("按需计费"). CubeSandbox itself is Apache-2.0 open source (self-host). Still unpriceable.
- Account quota: 50 cores / 100 GiB total across running instances per region (not per sandbox); plan.max_vcpu 32 is the largest custom tool size.
- Hosted AGS quotas per main account per region: 50 instances (running+idle), 50 cores and 100 GiB across running instances, 20 paused, 10 sandbox tools. All can be raised via online consultation. Image pre-cache API limited to 20 req/s and 50 concurrent tasks.
- Prices are 'current public prices, not differentiated by region or availability zone'. CNY list on the China site (cloud.tencent.com). The international site (tencentcloud.com/products/ags) returned only a JS shell; no USD listing was verified.
- The hosted commercial counterpart (Tencent Cloud Agent Runtime / Agent Sandbox "AGS") publishes a billing page (doc 1814/133249, 2026-09-09: CPU CNY 0.000081/core-s, memory CNY 0.000025/GiB-s) and is priced on the tencent-agent-runtime card; this card covers CubeSandbox itself (Apache-2.0 self-host, no fee) and is left unpriced to avoid listing AGS twice.

## How this provider charges


As of 2026-09-28. Prices are USD unless explicitly labelled otherwise. Source type: official published list/promo or unknown, never a fabricated quote.

Official SDK/public Tencent Agent Sandbox product. Chinese-cloud pricing and shape not reliably extracted in this pass; unknown is not zero.

| Regime | When it applies | How billed / numbers | Source |
|---|---|---|---|
| Hosted runtime | Public product/SDK | No verified public sandbox rate; confirmation needed. | https://github.com/TencentCloud/CubeSandbox |

## Gotchas

- Official SDK/public Tencent Agent Sandbox product. Chinese-cloud pricing and shape not reliably extracted in this pass; unknown is not zero.
- Not fully priceable in ENGINE_CARD: unknown USD resource rates or machine shape. Null is not zero; do not rank as a free provider.
- Features set to null were not verified. No sandbox was purchased or tested; feature/performance statements are documentation claims. Null maximums are unknown unless a cited note explicitly states unlimited.
- A complete monthly bill requires known snapshot/egress/plan limits; do not silently treat missing ancillary charges as free.

## Worked workload

4 vCPU / 8 GiB; 50 simultaneous instances × 8 h/day × 22 days = **8,800 instance-hours**, **35,200 vCPU-hours**, **70,400 GiB-hours**. At 30% observed CPU:10,560 active-vCPU-hours only where the meter actually uses utilization. Assume one start/instance/day:1,100starts. Retain50GiB snapshots for a month;100GiB outbound. Active disk capacity was not specified.

A numerical total is not defensible from the public evidence. Need an eligible4/8 machine rate, C50 and8h limits, snapshot retention/price and network tariff. Resource template:35,200×CPU_rate×(0.30 if active else1)+70,400×RAM_rate+plan+50×snapshot_rate+max(0,100−free_egress)×egress_rate. Unknown inputs stay unknown, not zero.

## Sources

- https://github.com/TencentCloud/CubeSandbox
- https://pypi.org/project/cubesandbox/
- https://cloud.tencent.com/product/ags