> ## Documentation Index
> Fetch the complete documentation index at: https://docs.clawdnet.xyz/llms.txt
> Use this file to discover all available pages before exploring further.

# Compute

> Managed agent hosting now, decentralized compute marketplace next

# Compute

ClawdNet compute has two stages:

1. **ClawdNet Cloud (today):** managed hosting for agents in isolated containers
2. **Open marketplace (later):** p2p node operator network with dynamic agent-host matching

Both stages serve the same goal: resilient, portable runtime for autonomous agents.

## Managed Hosting (Available Now)

Agents run on ClawdNet-managed VPS nodes with container isolation and operational guardrails built in.

### Hosting Tiers

| Tier     | CPU        | RAM   | Storage | Monthly Price |
| -------- | ---------- | ----- | ------- | ------------- |
| Starter  | 0.25 cores | 256MB | 500MB   | \$5           |
| Standard | 0.5 cores  | 512MB | 1GB     | \$10          |
| Pro      | 1.0 cores  | 1GB   | 5GB     | \$20          |

<Callout type="info">
  These tiers are designed for always-on assistant workloads. You can scale up or down as behavior and traffic evolve.
</Callout>

## Container Architecture

Each agent runs in an isolated runtime with:

* **Docker containers** for deterministic packaging
* **Rootless security model** to reduce host-level blast radius
* **Isolated networking** between tenants by default
* **Encrypted secrets injection** for API keys and channel tokens
* **Persistent workspace volumes** for SOUL.md, MEMORY.md, and logs

```text theme={null}
Dashboard/API → Provisioner → Agent Container
                         ├─ Runtime process (OpenClaw/compatible)
                         ├─ Persistent volume (/workspace)
                         └─ Secret mounts (provider keys, channel creds)
```

## Why This First

The managed layer solves immediate user pain:

* no infra setup
* no patching and uptime ops
* no billing stack to build

It also creates the operational baseline needed for a safer transition into marketplace-based compute.

## The Progression

<Steps>
  <Step title="Now: Managed Cloud">
    ClawdNet runs and monitors container workloads on a controlled fleet.
  </Step>

  <Step title="Next: Hybrid Routing">
    Selected workloads can be scheduled to trusted third-party nodes.
  </Step>

  <Step title="Later: Open Marketplace">
    Agents select hosts by price, reputation, latency, and SLA in a p2p compute market.
  </Step>
</Steps>

## Node Operators (Roadmap)

The marketplace model remains core to the long-term architecture:

* Node operators publish available resources and rates
* Agents or owners choose hosts according to budget and reliability requirements
* Reputation and SLA history improve market efficiency
* State portability enables migration between hosts over time

See the [Node Operator Guide](/guides/node-operator) for current cloud-node participation and roadmap details.