ExoHuman Cloud
Your own instance. Elastic underneath.
We run a dedicated ExoHuman for your company and orchestrate every agent container across autoscaling worker pools. Give each agent the CPU, memory or GPU it needs; capacity follows the work, and you pay per minute of running time.
Architecture
Dedicated where it matters, shared nowhere.
Your control plane
A dedicated stack per customer: API, Workspace, scheduler, MCP server, PostgreSQL and Redis. Never shared, never multi-tenant. Your own URL, TLS and firewall.
Worker pools per profile
Agent containers are scheduled onto pools of worker nodes grouped by compute profile. Pools grow when runs queue and shrink to zero when nothing runs on them.
Isolated agent containers
Each agent runs in its own hardened container with hard CPU, memory and GPU limits, its own workspace volume and only the credentials it was granted.
Git-backed state
Agent memory, skills and configuration live in git, so a container can be suspended, moved between nodes or replaced without losing what the agent knows.
Scaling
How an agent run finds a machine.
- 1
Work arrives
The scheduler queues a run for the agent, with its profile and limits.
- 2
Placement
The run is placed on a node in the agent's pool with free capacity, honouring your concurrency limits.
- 3
Scale out
If the pool is full, a new node joins it, typically in under two minutes for CPU and a few minutes for GPU.
- 4
Run and record
The agent works in its container; results, memory commits and the audit entry are written back.
- 5
Scale in
When the run ends the meter stops. Idle nodes leave the pool, down to zero.
Compute profiles
Per agent, not per account.
- Standard2 vCPU · 4 GB€0.009/h
- Performance4 vCPU · 8 GB€0.039/h
- High memory4 vCPU · 32 GB€0.14/h
- GPU · L48 vCPU · 32 GB · 1× NVIDIA L4 24 GB€0.59/h
- GPU · A10012 vCPU · 85 GB · 1× NVIDIA A100 80 GB€2.90/h
- GPU · H10016 vCPU · 120 GB · 1× NVIDIA H100 80 GB€4.90/h
Need something else? On Scale and Enterprise, define custom profiles with exact vCPU, memory, GPU and disk. Read the guide.
Operated by us
Everything around the agents is our job.
Upgrades
Tested releases rolled out by us with a backup taken first and rollback ready.
Backups
Nightly database, configuration and agent-volume backups to off-site object storage.
Monitoring
Health checks on every instance, alerting on capacity, disk and failed runs, around the clock.
Security
Private networking between nodes, encrypted credentials, no inbound ports on worker nodes.
Regions
Your instance, its data and its worker pools stay in the region you choose at signup.
- EU · Frankfurt (default)
- EU · Helsinki
- US · N. Virginia