SpaceScale
Currently in active development — follow our story

Infrastructure for
humans and agents.

Deploy isolated container workloads with Ignite. Create persistent Machines for humans and agents to use as real workspaces.

≡+ New shell
Linux shell▾
scale@edge:~$

One foundation. Two execution surfaces.

Ignite

Ignite runs services in microVMs. Bring an OCI image. SpaceScale starts it, routes it, observes it, and manages the runtime underneath.

Good for:

  • High-performance workloads
  • Long-running services

Machines

An isolated ready workspace where AI agents can work, run tasks, keep state, expose results, and hand work back to humans.

Good for:

  • Agent workspaces
  • Sandboxes
  • Preview environments

Performance at its peak

Fast starts. Direct routing. Predictable compute. SpaceScale is built to ignite workloads in milliseconds and keep traffic moving close to the metal.

Absolute Execution Isolation

Workloads are physically isolated at the CPU core level. No shared execution state between tenants. Dedicated reservations add CPU pinning for predictable performance.

We provide the endpoints,
you dictate the scale.

Bypass the dashboard and integrate SpaceScale into your own systems. Ignite workloads, start SpaceScale Machines, stream logs, and control lifecycle through the REST API for high-frequency human and agent workflows.

Example request
curl -X POST https://api.spacescale.io/v1/ignite \
  -H "Authorization: Bearer sk_live_..." \
  -H "Content-Type: application/json" \
  -d '{
    "image": "my-org/trading-api:latest",
    "regions": ["ca-east"],
  }'

Automatic Network Mapping

Our intelligent router detects exposed ports and maps each service to the right network path. Public services get TLS, DNS, and load balancing. Private services stay isolated on the internal mesh.

“Humans want to hand off workloads. Agents need environments they can actually control. SpaceScale is building both sides: Ignite for isolated container execution, and agent workspaces with shell access, files, exposed ports, logs, and lifecycle.”

SPACESCALE TEAM

INTUITIVE NATURE

Push once. Run everywhere.

Instant FailOver

Powered by a decentralized control plane. If a node fails, our router instantly redirects your users to healthy replicas at neural speed.

Global Edge Routing

Deploy once across regions. SpaceScale routes traffic to the nearest healthy physical node and gives web services, private services, and TCP workloads first-class network paths.

Zero-Drop Rollouts

Deploy updates without dropping active users. We wait for it to pass health checks, and seamlessly route new traffic to it naturally.

OBSERVABILITY

Live Telemetry & Logs

Zero ops does not mean zero visibility. Stream workload logs, lifecycle events, and VM telemetry in real time.

LOG STREAM

live
  1. EVENTnode joined fabric: ca-east-04anode joined fabric node=ca-east-04a region=ca-east capacity=32vcpu hypervisor=kvm
  2. INFOignite request accepted: trading-apiignite request accepted image=my-org/trading-api:latest target_regions=ca-east replicas=2
  3. INFOplacement auction opened: 9 biddersplacement auction opened auction=auc_7f21 region=ca-east bidders=9 workload=trading-api
  4. EVENTplacement auction complete: 12msplacement auction complete winner=ca-east-04a bid=12ms free_cpu=84% free_mem=71%
  5. INFOmicroVM ignition started: vm-fc-91b2microvm ignition started vm=vm-fc-91b2 jailer_uid=992 kernel=spacescale-kvm profile=ignite
  6. EVENTmicroVM ready: 118msmicrovm ready vm=vm-fc-91b2 boot_time=118ms workload=trading-api region=ca-east-04
  7. INFOstdout stream attached: scoutdruntime attached stdout stream vm=vm-fc-91b2 source=scoutd destination=control-plane
  8. EVENTrouter detected port: 8080/httprouter detected exposed_port=8080 protocol=http workload=trading-api visibility=public
  9. INFOendpoint mapped: trading-apiendpoint mapped domain=trading-api.spacescale.run target=vm-fc-91b2:8080 tls=managed
  10. WARNcapacity jitter delayed biddercapacity jitter delayed bidder node=ca-east-09c delay=50ms reason=mid_utilization
  11. INFOhealth probe registered: /healthzhealth probe registered path=/healthz interval=10s timeout=2s vm=vm-fc-91b2
  12. EVENTreplica launched: ca-east-09breplica launched vm=vm-fc-92a8 node=ca-east-09b router_weight=50 lifecycle=active
  13. INFOrequest routed: edge ca-eastrequest routed edge=ca-east route=trading-api.spacescale.run latency=4ms upstream=vm-fc-91b2
  14. EVENTworkload live: healthyworkload live url=https://trading-api.spacescale.run replicas=2 logs=streaming status=healthy
  15. INFOnode heartbeat: ca-east-04anode heartbeat received node=ca-east-04a microvms=18 alloc_cpu=62% alloc_mem=49%
  16. EVENTfabric route installedprivate fabric route installed workload=trading-api cidr=10.42.18.0/24 mesh=ca-east
  17. WARNauction bidder skippedauction bidder skipped node=ca-east-02d reason=reserved_core_pressure available_cores=1
  18. INFOlog checkpoint persistedlog stream checkpoint persisted workload=trading-api cursor=log_19fd8 consumer=dashboard
  19. EVENTlifecycle: starting -> runninglifecycle state transition workload=trading-api previous=starting current=running source=scaled
  20. INFOcontrol-plane ack: committedcontrol-plane ack received auction=auc_7f21 placement=committed reservations=2 duration=134ms
  21. EVENTnode joined fabric: ca-east-04anode joined fabric node=ca-east-04a region=ca-east capacity=32vcpu hypervisor=kvm
  22. INFOignite request accepted: trading-apiignite request accepted image=my-org/trading-api:latest target_regions=ca-east replicas=2
  23. INFOplacement auction opened: 9 biddersplacement auction opened auction=auc_7f21 region=ca-east bidders=9 workload=trading-api
  24. EVENTplacement auction complete: 12msplacement auction complete winner=ca-east-04a bid=12ms free_cpu=84% free_mem=71%
  25. INFOmicroVM ignition started: vm-fc-91b2microvm ignition started vm=vm-fc-91b2 jailer_uid=992 kernel=spacescale-kvm profile=ignite
  26. EVENTmicroVM ready: 118msmicrovm ready vm=vm-fc-91b2 boot_time=118ms workload=trading-api region=ca-east-04
  27. INFOstdout stream attached: scoutdruntime attached stdout stream vm=vm-fc-91b2 source=scoutd destination=control-plane
  28. EVENTrouter detected port: 8080/httprouter detected exposed_port=8080 protocol=http workload=trading-api visibility=public
  29. INFOendpoint mapped: trading-apiendpoint mapped domain=trading-api.spacescale.run target=vm-fc-91b2:8080 tls=managed
  30. WARNcapacity jitter delayed biddercapacity jitter delayed bidder node=ca-east-09c delay=50ms reason=mid_utilization
  31. INFOhealth probe registered: /healthzhealth probe registered path=/healthz interval=10s timeout=2s vm=vm-fc-91b2
  32. EVENTreplica launched: ca-east-09breplica launched vm=vm-fc-92a8 node=ca-east-09b router_weight=50 lifecycle=active
  33. INFOrequest routed: edge ca-eastrequest routed edge=ca-east route=trading-api.spacescale.run latency=4ms upstream=vm-fc-91b2
  34. EVENTworkload live: healthyworkload live url=https://trading-api.spacescale.run replicas=2 logs=streaming status=healthy
  35. INFOnode heartbeat: ca-east-04anode heartbeat received node=ca-east-04a microvms=18 alloc_cpu=62% alloc_mem=49%
  36. EVENTfabric route installedprivate fabric route installed workload=trading-api cidr=10.42.18.0/24 mesh=ca-east
  37. WARNauction bidder skippedauction bidder skipped node=ca-east-02d reason=reserved_core_pressure available_cores=1
  38. INFOlog checkpoint persistedlog stream checkpoint persisted workload=trading-api cursor=log_19fd8 consumer=dashboard
  39. EVENTlifecycle: starting -> runninglifecycle state transition workload=trading-api previous=starting current=running source=scaled
  40. INFOcontrol-plane ack: committedcontrol-plane ack received auction=auc_7f21 placement=committed reservations=2 duration=134ms

VM TELEMETRY

CPU
23%
MEMORY
4.2 / 16 GB
DISK I/O
120 MB/s read
NETWORK IN
840 Mbps
NETWORK OUT
320 Mbps
UPTIME
14d 6h 32m

LIFECYCLE EVENTS

  1. VM booted (ca-east-04)
  2. VM booted (ca-east-09)
  3. Health checks registered
  4. TLS provisioned (ACME)
  5. L4 Router configured
  6. DNS A records live
  7. First request received
  8. Auto-scaled replica 2
  9. Cert renewed (auto)
  10. Config hot-reload

ALMOST READY...

We are in active development

Read our engineering blog to learn more about our design decisions, architecture tradeoffs, and the systems work behind SpaceScale as we build toward the first public runtime.

Read how we're building it