COLLIMATE
Docs Blog Get a key Contact
01/Pricing · Fork · Schedule · Store

Fast live-fork microVM
sandboxes for agents & RL.

Prepare a sandbox once, fork it many; each clone keeps live process memory. Start free with a demo key, no card. Every tier runs the same hardware-isolated microVMs; plans differ only in how much you can run at once and for how long.

agent@collimate — api.collimate.ai 
$ curl -sX POST api.collimate.ai/v1/sandboxes \
    -H "Authorization: Bearer $COLLIMATE_API_KEY" -d '{"template":"python"}'
{ "id": "sbx_a1f09c", "state": "running" }

$ curl -sX POST api.collimate.ai/v1/sandboxes/sbx_a1f09c/exec \
    -d '{"command":"pytest -q"}'   # run to a decision point
{ "exit_code": 0, "stdout": "14 passed in 2.31s" }

$ curl -sX POST api.collimate.ai/v1/sandboxes/sbx_a1f09c/fork \
    -d '{"count":8}'   # branch the live VM · fork ~2.4 ms p50
{ "parent": "sbx_a1f09c", "children": [ ... ] }  # 8 live clones, memory preserved

$ curl -sX POST api.collimate.ai/v1/sandboxes/sbx_a1f09c/suspend  # stops the meter
{ "id": "sbx_a1f09c", "state": "suspended" }
02/Three tiers · one engine
Demo
Freeno card

Self-serve. Kick the tires on real microVMs.

  • 2 concurrent sandboxes, 1 hour each
  • Live-fork, exec, suspend / resume: every verb
  • Built-in python template
  • No card required
  • Community support
Your own demo key DEMO · concurrency 2 · 1h

Sign up with an email and the gateway mints a key scoped to your tenant — rate-limited & capped per key. Shown once, right after signup.

Get your demo key→
Most teams
Pro
Usageno platform fee

Production fan-out on api.collimate.ai.

  • Up to 100 concurrent sandboxes, 24h sessions
  • Per-tenant keys, private templates & lineage
  • Bring-your-own image (self-hosted today)
  • Live-fork for tree search & RL rollouts
  • Metered per-second usage; suspend stops the meter
  • Direct support from the engineers
Start on Pro→
Enterprise
Custominvoice

Dedicated capacity, SSO, or run it in your cloud.

  • Unlimited concurrency & session length
  • Dedicated nodes, no shared hosts
  • SSO, per-tenant key tiers & limits
  • Bring-your-own-cloud: data never leaves your VPC
  • Onboarding, SLA & security review
Talk to sales→
03/Compute rates

Pay for compute, not for sandboxes.

Pro is billed per second on the CPU and memory a sandbox holds while it is live, plus snapshot storage while it is suspended. Spawning a sandbox and forking one are free, so fan-out never becomes a line item. Suspend a sandbox and the meter stops.

vCPU

$0.040 / vCPU-hour, billed per second while running.

Memory

$0.012 / GB-hour, billed per second while running.

Suspended storage

$0.10 / GB-month while a sandbox is suspended.

Spawn & fork

Free. No charge to spawn a sandbox or fork a live one.

A 1 vCPU, 2 GB sandbox costs about $0.064 / hour while live. Demo usage is metered for observability but priced at $0.

04/What every tier shares

Same engine, top to bottom.

Every tier runs the exact same Firecracker microVMs on the same fleet. We never sell you a weaker sandbox: the Demo key and an Enterprise key hit identical isolation and identical fork latency. Tiers only change how much you can run at once, how long, and how you're supported. Metering is per-second while a sandbox is live; suspend a sandbox and the meter stops.

Fork

Fork a running microVM in single-digit milliseconds. The clone keeps live process memory. Prepare once, branch many.

Schedule

Warm-pool and affinity placement keep sandboxes off your trainer's critical path, built for agentic fan-out and RL rollouts alike.

Store

Suspend to snapshot storage between turns; resume with memory intact. You pay per-second live, per-GB-month suspended.