Choosing a deployment

Every ExoHuman customer gets a dedicated, isolated instance on ExoHuman Cloud — never a shared tenant. What you choose is where the agents run: on our elastic pools, on machines you connect as worker nodes, or a mix of both. Governance — approvals, audit, budgets, guardrails — is identical everywhere.

At a glance

| | Cloud pools | Your worker nodes | |---|---|---| | Infrastructure | Hosted, scaled and updated by us | Machines you own and control | | Isolation | Isolated containers on our fleet | Containers inside your perimeter | | Agent compute | Elastic worker pools, autoscaled per profile, scale to zero when idle | Your servers, VMs and GPUs; capacity you already have | | GPU | L4, A100 and H100 profiles on demand | The GPUs you already own | | Operations | We provision, monitor and back up pools | You keep the machine healthy; the instance handles placement | | Billing | Per-minute compute while an agent runs | Included — it is your hardware | | Time to connect | Nothing to do | Two steps: create the node, paste one command |

Choose cloud pools when

• your workload is spiky — agents that run a few minutes an hour cost a few minutes an hour;
• you need GPUs occasionally and do not want to buy them;
• you want nothing to operate.

Start at https://exohuman.cc/signup or read the Quick start.

Choose your worker nodes when

• policy or contracts require an agent’s working data to stay inside your own network;
• you already own capacity, especially GPUs, that you want agents to use;
• a specific machine holds data or tools an agent must reach locally.

Connecting a node takes two steps and is self-service — see Worker nodes.

Mixing both

Most teams start fully on cloud pools and connect their own nodes later for specific agents — a GPU box, a machine that can reach an internal system, a regulated dataset. Placement follows the labels you give each node, so the right agents land on the right hardware automatically.