# Lizard

Official pricing: https://lizard.build/pricing  
Category: agent-sandbox · Isolation: container

## Pricing regimes (raw)

- **Sandbox (4 vCPU / 4 GiB, flat per second while it runs)** (sizes): medium 4 vCPU/4 GiB $0.018/h

## Features

Yes: persistent disk, what survives a stop, volumes, ≥24 h sessions, code interpreter, preinstalled agents, HTTPS preview URLs, open internet, Python, EU data residency, extra volumes

No: snapshots, memory snapshots, fork/clone, pause/resume, whole disk survives a stop, idle auto-stop, custom image (Docker or snapshot), start from your own snapshot, your own Docker/OCI image, full VM (own kernel), nested virtualization, computer-use API, browser + desktop control, SSH, public IPv4, raw TCP inbound, static egress IP, self-hosting, BYOC, open source, SOC 2, HIPAA, SSO, GPU, Windows, macOS, wake on request, live resize, live fork (no pause), memory fork, audit logs, automatic snapshots, snapshots on demand, hosted agent API (their own agent), gVisor or VM (no shared kernel), shared volumes

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

## Company

- Ownership: independent
- Stage: unknown
- Accelerators: none found
- Total raised: unknown
- Revenue: unknown

Lizard (lizard.build) is an AI-native cloud deployment platform (apps, managed Postgres/Redis, object storage, sandboxes) aimed at being driven by AI coding agents. No public information on funders, investors, funding stage/amount, accelerator participation, or revenue was found: the marketing site, GitHub org (lizard-build, 11 repos), LinkedIn company page, and X/Discord links carry no financial disclosures, and searches (DuckDuckGo, Bing, HN Algolia) for "lizard.build" funding/seed/Series A/Y Combinator returned no relevant hits within the ~15 minute research budget. Could not confirm whether it is bootstrapped, VC-backed, or a side project.

## Caveats

- Container isolation: sandboxes are Kubernetes pods on runc sharing the host kernel, not VMs.
- The landing page advertises SSH, pause/resume, snapshot/restore, a desktop and a Claude Code / Codex prompt command; the SDK reference says SSH is not available, that pause, resume, fork, snapshot and restore return HTTP 501, and lists no desktop template or prompt method.
- Sizes: the pricing page sells Small 2 vCPU / 4 GB ($0.009/h), Medium 4 / 8 ($0.018/h) and Large 8 / 16 ($0.036/h), but the SDK reference says current sandboxes have fixed limits of 4 vCPU and 4096 MiB that create options don't change. Priced here as Medium at what the docs say you get: 4 vCPU / 4 GiB.
- The flat sandbox rates are about a tenth of Lizard's own per-resource rates (CPU $0.025/vCPU-h, memory $0.0125/GB-h).
- Self-serve limit is 100 sandboxes at once (after $20 paid in total); more needs Enterprise. The landing calculator goes up to 2,000 at once.
- Persistent Volumes are node-local and attach to one sandbox at a time.
- Public usage counter (lizard.build/api/public/stats, 2026-10-01): 436 workspaces, 8,870 sandbox starts in total.

## How this provider charges


Public product verified from pricing, SDK and sandbox documentation. Discovery: [DEV, September 23 2026](https://dev.to/yuraoak/e2b-sandboxes-pricing-lifecycle-and-alternatives-m0h); this establishes presence, **not original launch date**. The [public SDK repository](https://github.com/lizard-build/lizard-sdk) is also available. All prices USD list unless labeled trial.

| Regime | When it applies | How billed | Numbers | Source |
|---|---|---|---|---|
| Running sandbox | Fixed public 4-vCPU/4096-MiB template | Measured CPU and RAM seconds | CPU $0.000006948/vCPU-s = $0.0250128/vCPU-h; RAM $0.000003474/GB-s = $0.0125064/GB-h | [Pricing](https://lizard.build/pricing), [actual size constraint](https://lizard.build/docs/sandboxes.md) |
| Paused guest | vCPUs frozen; state in host RAM | No on-demand compute; retained attached volumes separate | $0 compute while paused; retention unsafe across host failure or expiration issue | [Dashboard](https://lizard.build/docs/sandboxes/dashboard.md), [known issues](https://lizard.build/docs/platform/known-issues.md) |
| Persistent volume | Created storage, including after sandbox ends | GB-seconds | $0.000000054/GB-s = $0.141912/GB per 730-hour month | [Pricing](https://lizard.build/pricing), [volumes](https://lizard.build/docs/sandboxes/volumes.md) |
| Network | Outbound bytes from first byte | GB transferred | $0.045/GB; ingress and per-request fee $0 | [Pricing](https://lizard.build/pricing) |
| PAYG account | Standard accounts | No base/seat subscription | $0 recurring base; purchased credits do not expire | [Pricing](https://lizard.build/pricing) |
| Trial | New account | Promotional usage credit | $10, expires after 31 days | [Pricing](https://lizard.build/pricing) |
| Enterprise | Custom infrastructure/compliance contract | Sales-priced | Unknown; do not assume public unit rates or capacity | [Pricing](https://lizard.build/pricing) |

## Gotchas

- **4 vCPU / 8 GiB is not publicly supported.** The sandbox docs now explicitly fix the create API to 4 vCPU/4096 MiB. App quotas of 20 vCPU/40 GB do not override that restriction. [Sandbox specification](https://lizard.build/docs/sandboxes.md).
- Marketing suggests custom templates and shared storage, but current docs expose only `base` and `code-interpreter-v1`, with no public template-upload flow. A volume attaches to one sandbox at a time and stays on its node. Prefer the specific docs to generic marketing. [Templates](https://lizard.build/docs/sandboxes.md), [volume lifecycle](https://lizard.build/docs/sandboxes/volumes.md).
- Pause is **not durable snapshot storage**. RAM stays in host memory. A documented lifecycle issue may expire a paused VM at its original deadline. Default SDK lifetime is five minutes; raw `timeoutMs: 0` disables expiration. [Known issues](https://lizard.build/docs/platform/known-issues.md), [quickstart](https://lizard.build/docs/sandboxes/quickstart.md).
- Pricing says measured memory, not allocated memory; the precise measurement statistic is not documented. Active RAM is approximated by estimator `ram_basis=active`, not asserted to be resident-set sampling with a known interval. [Pricing](https://lizard.build/pricing).
- Payment processing fees and top-up minimum appear only at checkout. Do not assume $0. [Pricing](https://lizard.build/pricing).

## Required worked example

Workload: 4 vCPU / 8 GiB, 50 simultaneous sandboxes × 8 hours/day × 22 days = **8,800 sandbox-hours**; CPU utilization 30%; retained snapshots 50 GiB-month; egress 100 GiB.

**Result: unsupported / no valid total.** A documented sandbox has only 4 GiB RAM, capacity for 50 is not confirmed, and durable VM snapshots are not offered. The 4/8 arithmetic `$0.2001024/h` at full utilization is deliberately not encoded as a purchasable quote.

For arithmetic validation only, changing the request to **4 vCPU / 4 GiB** and assuming all 4 GiB are billable measured memory yields:
- CPU: `8,800 × 4 × 0.30 × 0.0250128 = $264.135168`.
- Memory: `8,800 × 4 × 0.0125064 = $440.225280`.
- Compute subtotal: **$704.360448**; egress: `100 × 0.045 = $4.50`.
- An optional 50-GiB volume for 730 hours would add **$7.0956**, but is **not equivalent to 50 GiB of retained snapshots** and cannot be shared simultaneously across all 50 guests.

No trial credit deducted; payment fees and capacity approval remain unknown. This altered workload is not a quote for the required example.