High Concurrency Sandboxes in EU & US, Fast and Affordable.
San Francisco, CA
Joined February 2025
- Tweets180
- Following47
- Followers289
- Likes114
Pinned Tweet
Some users asked where is comparison with @boatdotdev. Added it, and scaled the bill to a real agent fleet
100 small sandboxes, 8 hours a day, 30 days:
> Lizard $389
> Boat $432
> Cloudflare $1,673
> Vercel $2,650
> E2B $3,974
> Daytona $3,974
> Modal $5,712
lizard.build/boat
Very close, but Lizard still better price!
This list is missing the sandbox that costs $11.66 a month. Modal charges $171 for the same box
Same small sandbox, running 24/7 for 30 days:
> Lizard $11.66
> Cloudflare $50
> Vercel $79
> E2B $119
> Daytona $119
> Modal $171
lizard.build/options
Firecracker microVMs, 228 ms cold start, snapshots and pause built in. First week free
This list is missing the sandbox that costs $11.66 a month. Modal charges $171 for the same box
Same small sandbox, running 24/7 for 30 days:
> Lizard $11.66
> Cloudflare $50
> Vercel $79
> E2B $119
> Daytona $119
> Modal $171
lizard.build/options
Firecracker microVMs, 228 ms cold start, snapshots and pause built in. First week free
Most goated alternatives to Vercel sandbox:
1. E2B (@e2b) - e2b.dev
2. Instabox (@insforge) - instabox.sh
3. Modal sandbox(@modal ) - modal.com
4. Sandboxer(by @jonoringer) - sandboxer.com
5. Daytona(@daytonaio ) - daytona.io
6. Cloudflare Sandbox(@Cloudflare ) - cloudflare.com/products/sand…
There is. It's metal.
Software get copied in a week. Hardware, power and servers you own don't.
Get a nice long term spot in sandbox industry:
lizard.build/moat
700 VMs of minecraft is cool. we brought the shotgun: 50 Lizard sandboxes running doom
One sandbox first: 4 vCPU, 8 GB, a desktop, Doom installed. It was up in under a second. We snapshotted it in 18 seconds, then restored that snapshot 50 times
lizard.build/launch
Every copy loaded its own map and its own bot, with a real screen recording from each desktop. No mockups, the grid in the video is all 50
> 50 sandboxes, 4 vCPU / 8 GB each
> one snapshot, 50 different games
> $0.0324/h per sandbox, $1.62/h for the whole grid
> desktop, snapshots, pause and resume built in
Next time we'll let the agents play
nice try
we spun up 700 boat.dev VMs 4c x 8gb, started codex + minecraft, made it all play in multiplayer together
they all booted from a snapshot in the first 10s, which cost $0.07, same price per VM, core & ram
but much more impressive capabilities and scale
Railway's report is out: rotate every secret stored before Oct 21!
If you're rotating everything anyway, set the new keys on Lizard and let your coding agent move the rest. Apps, workers, Postgres, Redis, S3 and your agent sandboxes, on one bill
lizard.build/rotate
> npx skills add lizard-build/skill -g
> tell your agent "migrate us to Lizard"
> it recreates services, databases and the data inside
Agent sandboxes come out 18x cheaper: 10 running 24/7 cost $118/mo on Lizard vs ~$2,129 on Railway. First week free
The preliminary report related to the October 21st Security Incident is now available.
blog.railway.com/p/prelimina…
For anyone moving off Railway this week
10 sandboxes (2 vCPU, 4 GB) running 24/7 for a month:
> Lizard: $118
> Railway: ~$2,129
lizard.build/move
> app hosting costs less too: $18.26 per vCPU and $9.13 per GB of RAM a month
> Postgres, Redis, S3 and agent sandboxes on one bill
> tell your coding agent "migrate us to Lizard" and it moves your apps, databases and data
Rotate your secrets first, then move. First week free
An urgent notice from the Railway team
blog.railway.com/p/security-…
Railway deploys still paused? Here's a Railway alternative
Rotate your secrets first, then tell your coding agent "migrate us to Lizard": it redeploys your apps and workers, sets up Postgres, Redis and S3 and imports your data
lizard.build/unpause
> set fresh secrets on Lizard, don't copy the old ones
> apps, databases and agent sandboxes on one account, one bill
> first week free
The easy way to move is right here 👇
Deploys paused and secrets to rotate? Your agent can move you to Lizard sandboxes and hosting in one prompt
If your platform froze deploys this week, you don't have to wait it out. Rotate your secrets first, then hand the move to your coding agent
lizard.build/column
> install the Lizard Skill: npx skills add lizard-build/skill -g
> tell your agent "migrate us to Lizard": it redeploys your services and workers, provisions Postgres, Redis and S3 and imports your data
> set your new, rotated secrets on Lizard instead of copying the old ones
> point your domain and you're back to shipping
Works with Claude Code, Codex, Cursor and any agent with a terminal. Your agents can keep building in Lizard sandboxes, and the first week is free
Deploys paused and secrets to rotate? Your agent can move you to Lizard sandboxes and hosting in one prompt
If your platform froze deploys this week, you don't have to wait it out. Rotate your secrets first, then hand the move to your coding agent
lizard.build/column
> install the Lizard Skill: npx skills add lizard-build/skill -g
> tell your agent "migrate us to Lizard": it redeploys your services and workers, provisions Postgres, Redis and S3 and imports your data
> set your new, rotated secrets on Lizard instead of copying the old ones
> point your domain and you're back to shipping
Works with Claude Code, Codex, Cursor and any agent with a terminal. Your agents can keep building in Lizard sandboxes, and the first week is free
An urgent notice from the Railway team
blog.railway.com/p/security-…
Migrate on Lizard easily via CLI
Builds are temporarily paused
We are currently investigating this issue
Status: Investigating status.railway.com/incident/…
We timed 1,200 sandbox boots. Lizard's median: 238 ms
300 cold starts each on Lizard, E2B, Daytona and Modal. Every ball in the video is one sandbox boot, dropped into a bin by how long it took to start
lizard.build/session
Median cold start:
> Lizard 238 ms
> E2B 270 ms
> Daytona 796 ms
> Modal 865 ms
Lizard had the lowest median and average of the four. When your agent spins up a sandbox for every task, that wait adds up fast. The full table, up to p99, is on our sandboxes page
Your agents sat idle all weekend. You paid for every hour of it
Most agent setups keep their machines running around the clock, even when nobody gives them work. 50 coding agents on Small sandboxes running 24/7 cost $136 a week. Paused outside working hours, the same 50 cost $36
lizard.build/replay
> a paused Lizard sandbox costs $0 an hour
> resume on Monday with the same memory, files and running processes
> billed per second while it runs, so a 10-minute task costs 10 minutes
> apps and databases on Lizard go to sleep on their own after 10 minutes with no requests
Pause your sandboxes when the work stops. Your agents won't notice, your bill will
Your agents sat idle all weekend. You paid for every hour of it
Most agent setups keep their machines running around the clock, even when nobody gives them work. 50 coding agents on Small sandboxes running 24/7 cost $136 a week. Paused outside working hours, the same 50 cost $36
lizard.build/replay
> a paused Lizard sandbox costs $0 an hour
> resume on Monday with the same memory, files and running processes
> billed per second while it runs, so a 10-minute task costs 10 minutes
> apps and databases on Lizard go to sleep on their own after 10 minutes with no requests
Pause your sandboxes when the work stops. Your agents won't notice, your bill will
Your agent moved to a new sandbox mid-task. It didn't notice
Pause it in the EU, restore it in the US 300-900 ms later, and it picks up on the same line with the same memory, processes and files. Fork it mid-task and the copies keep everything too
Try it free for a week: lizard.build/sandboxes/resum…
> pause and resume with memory and running processes kept
> restore from a snapshot on any machine, EU or US
> fork a running agent and let the copies try different ideas
> a paused sandbox costs $0, a running one is billed per second
What would this cost for your agents? There's a calculator on the page
1,000 deploys from a sandbox at 3 am. Your agents fixed every failed one
Your coding agent builds in a Lizard sandbox and ships straight from it. When a deploy breaks, the CLI doesn't just print a wall of logs: every error comes back as structured JSON with a reason and an exit code, so the agent in the sandbox reads why it failed and fixes it on its own
lizard.build/sandboxes/guard
In our example the agent:
> saw the health check fail with "DATABASE_URL is not set" and added Postgres with lizard add postgres
> redeployed, saw "table users does not exist" and ran the migrations
> shipped on the third try, with zero messages to you
Every sandbox is its own isolated microVM. Works with Claude Code, Codex, Cursor and any agent with a terminal
1,000 deploys from a sandbox at 3 am. Your agents fixed every failed one
Your coding agent builds in a Lizard sandbox and ships straight from it. When a deploy breaks, the CLI doesn't just print a wall of logs: every error comes back as structured JSON with a reason and an exit code, so the agent in the sandbox reads why it failed and fixes it on its own
lizard.build/sandboxes/guard
In our example the agent:
> saw the health check fail with "DATABASE_URL is not set" and added Postgres with lizard add postgres
> redeployed, saw "table users does not exist" and ran the migrations
> shipped on the third try, with zero messages to you
Every sandbox is its own isolated microVM. Works with Claude Code, Codex, Cursor and any agent with a terminal
96 agents, one shared bucket, and a city built together
We gave 96 agents their own Lizard sandboxes, one shared bucket and a private network, and let them build a city together. Every agent reads and writes the same bucket and talks to the others directly, even between our EU and US regions
lizard.build/sandboxes
On most clouds that's where the bill hides: a fee on every read and write, and a fee on every message between machines. On Lizard both are $0
> Object Storage has no per-request fees
> traffic between sandboxes runs on a private network, it's free and doesn't count toward your egress
> each agent runs in its own isolated microVM, $0.0162 an hour for a Small sandbox
You pay for the sandboxes while they run and for what you store. Not for your agents talking to each other
You shipped an update. 1,000 users lost their files. Not on Lizard
Every app that gives users their own agent hits this sooner or later: you push a new version, rebuild the environments, and whatever your users made inside them is gone
On Lizard each user gets their own sandbox, and their files live on a persistent volume at /workspace, not inside the sandbox. Ship a new version, snapshot it once, and every user's sandbox comes back on it in 1 to 3 seconds with the same volume attached
lizard.build/sandboxes
> the sandbox is new, the user's files stay exactly where they were
> no migration scripts, no copying data between machines
> each user's sandbox is its own isolated microVM
> billed per second while it runs, and a paused sandbox costs nothing
Ship as often as you want. Your users won't notice anything except the new features
You shipped an update. 1,000 users lost their files. Not on Lizard
Every app that gives users their own agent hits this sooner or later: you push a new version, rebuild the environments, and whatever your users made inside them is gone
On Lizard each user gets their own sandbox, and their files live on a persistent volume at /workspace, not inside the sandbox. Ship a new version, snapshot it once, and every user's sandbox comes back on it in 1 to 3 seconds with the same volume attached
lizard.build/sandboxes
> the sandbox is new, the user's files stay exactly where they were
> no migration scripts, no copying data between machines
> each user's sandbox is its own isolated microVM
> billed per second while it runs, and a paused sandbox costs nothing
Ship as often as you want. Your users won't notice anything except the new features
Your agents can talk to each other, and the traffic is free
Lizard sandboxes reach each other over a private network, even across regions. Put a planner in the EU and its workers in the US, pass files, diffs and test results between them, and pay $0 for that traffic
lizard.build/sandboxes
> private network between sandboxes, not the public internet
> works between EU and US regions
> traffic between sandboxes is free and doesn't count toward your 2 TB of egress
> every agent runs in its own isolated microVM
One bug, 50 agents, first fix wins
Set up a sandbox once and snapshot it, or fork one that's already running with its memory and processes. Each copy is ready in 1 to 3 seconds. Give 50 agents the same bug and a different idea each, and keep the first fix that turns the tests green
lizard.build/sandboxes
> 50 Small sandboxes cost $0.81 an hour while they run
> pause the other 49 the moment one passes, and they cost nothing
> every fork runs in its own isolated microVM
One bug, 50 agents, first fix wins
Set up a sandbox once and snapshot it, or fork one that's already running with its memory and processes. Each copy is ready in 1 to 3 seconds. Give 50 agents the same bug and a different idea each, and keep the first fix that turns the tests green
lizard.build/sandboxes
> 50 Small sandboxes cost $0.81 an hour while they run
> pause the other 49 the moment one passes, and they cost nothing
> every fork runs in its own isolated microVM
Lizard retweeted
Run 100 coding agents at once
Every Lizard sandbox boots in about 238 ms, and Pro runs up to 100 of them at the same time. Give each one Claude Code or Codex and its own task, and all 100 work in parallel
lizard.build/sandboxes
> each agent in its own isolated microVM
> 100 Small sandboxes cost $1.62 an hour together
> billed per second, so a 20-minute task costs 20 minutes
> pause them when they're done and they cost nothing until you resume
Run 100 coding agents at once
Every Lizard sandbox boots in about 238 ms, and Pro runs up to 100 of them at the same time. Give each one Claude Code or Codex and its own task, and all 100 work in parallel
lizard.build/sandboxes
> each agent in its own isolated microVM
> 100 Small sandboxes cost $1.62 an hour together
> billed per second, so a 20-minute task costs 20 minutes
> pause them when they're done and they cost nothing until you resume
Your agent reads files all day. Here's what that costs
We priced one month of a typical agent workload: 1 TB stored, 100 GB served, 500M reads and 50M writes
Lizard: $18
Cloudflare R2: $420
AWS S3: $473
lizard.build/sandboxes
> no per-request fees on Lizard Object Storage, you pay for storage and egress only
> storage at $0.0135 per GB-month, S3-compatible
> sits next to your sandboxes, apps and databases on one bill
Competitor prices are from their public pricing pages as of October 2026