What people don't realize: your agent needs 5 different clouds just to ship one app lizard.build/launch Sandboxes on one, CI on another, then hosting, Postgres and storage somewhere else, each with its own bill and limits. Lizard runs the whole loop in one place: > sandboxes from $0.0162 an hour, $0 while paused > GitHub Actions runners at $0.00027 a minute > apps billed on the CPU they actually use, $18.26 per vCPU a month > Postgres, Redis and S3 on the same bill > no seat fees, add the whole team Try it free for 7 days

Good catch. Everyone except Vercel and Cloudflare bills every running second, so we gave those two their best case, 10% busy. At full load it's about 290 hours on Vercel and 500 on Cloudflare. Lizard stays at 6,111 either way

Fair question. Every number is $99 divided by each provider's hourly price for 2 vCPU / 4 GB, from their own pricing pages: Lizard $0.0162/h Boat $0.018/h E2B and Daytona $0.1656/h Modal $0.238/h Vercel and Cloudflare bill CPU only while it's busy, so we priced them at 10% CPU. At full load they cost more and get fewer hours

We ran 10 agents on one product: one built it, nine reviewed it. Coordination broke well before the prompts did

Hours-long runs also need a machine that won't expire halfway through

Paying only while it works should be the default for agent machines

Nice to see a public benchmark. How much of the 19.4s is boot vs install?

How much CPU does the $1 box get? Curious how it holds up when an agent runs a full test suite

Which matters most, the subs or the multiplayer? Running Claude Code and Codex in a cloud box is the easy part now, shared threads are where most tools fall short

Depends what your agents run. If it's untrusted code, pick something with a real VM per sandbox, not a container. We built Lizard on Firecracker for that reason

Your Conductor plan is gone, but your agents still need sandboxes? $99 buys 6,111 hours on Lizard lizard.build/pricing Small sandbox hours (2 vCPU / 4 GB) for $99: > Lizard 6,111 > Boat 5,500 > Cloudflare 1,447 > Vercel 897 > E2B 598 > Daytona 598 > Modal 416 On our $99 plan sandbox starts are unlimited. Try Lizard free for 7 days and upgrade when your team needs more

Can your AI agent survive its own death? On Lizard it keeps up to 500 GB of files after its sandbox dies lizard.build/volume Lizard Persistent Volumes are NVMe SSDs that live outside the sandbox: > plug one in and your agent saves its files to it > kill the sandbox, the disk and the files stay > start a new sandbox with the same disk, and the agent picks up where it stopped > you pay only for the space you use, $0.14 per GB a month

Exactly. Writing code is becoming the easy part; getting it running, tested, and shipped is the new bottleneck. That’s the problem we’re building Lizard to solve.

What means better ci for you?

@dodeja , if you're paying $1k a month for CI, the same builds fit in a $99 plan on Lizard lizard.build/optimise Lizard runs your GitHub Actions jobs in our Sandboxes 250,000 build minutes a month on 2 vCPU runners with no limit on starts: > Blacksmith $1,000 > Lizard $99 Our Pro 5x plan is $99 a month and includes $99 of usage. At $0.00027 a minute that covers about 366,000 build minutes and every job runs in its own Firecracker VM

terrible

we have volumes big volumes

🦎

Where is Lizard? 🦎

we proved it!