AgentPuter Documentation

Managed hosting for OpenClaw. Your AI assistant, always on.

Ship your first agent in 2 minutes

OpenClaw pre-installed. Channels via Web UI. No DevOps.

Start Quick Start →
🚀

Get Started

Create a Pod with OpenClaw pre-installed.

⚙️

Configure

Customize your Pod and environment.

🔒

Understand

Security model, costs, and troubleshooting.

Skills & ClawHub

OpenClaw includes 50+ built-in skills and 565+ community skills via ClawHub.
AgentPuter pre-loads popular skills and lets you manage them from the Dashboard.

$ agentputer skills list

Built-in: 50 skills loaded

ClawHub: 565+ available

Run: agentputer skills install <name>

What you'll build:

your-terminal

$ agentputer create my-agent

✓ Pod ready (OpenClaw v2026.2.6 pre-installed)

✓ Gateway running (ws://pod-abc123:18789)

$ agentputer channels add telegram

✓ Bot @my_agent_bot connected

$ agentputer status

✓ Pod: running | OpenClaw: v2026.2.6 | Channels: 1

✓ Skills: 12 active | ClawHub: connected

🔄

Coming from self-hosted OpenClaw?

Migration guide coming soon. In the meantime, Quick Start works for you too.

CONFIGURE

Pod Settings

A Pod is your AgentPuter server instance — an always-on container running OpenClaw. Configure region, resources, and storage from the Dashboard.

$ agentputer pod info my-agent

Region: us-east-1

CPU: 2 vCPU | RAM: 2 GB

Storage: 40 GB SSD (2.1 GB used)

OpenClaw: v2026.2.6 (auto-update: on)

Status: running (uptime: 14d 6h)

Environment & Secrets

Configure your AI model API keys and other secrets. AgentPuter stores them encrypted and injects them into your OpenClaw instance — no need to SSH in and edit JSON files.

$ agentputer config set agent.model=anthropic/claude-opus-4-6

✓ Model updated

$ agentputer secrets set ANTHROPIC_API_KEY=sk-ant-...

✓ Secret stored (encrypted)

Recommended model: Anthropic Claude Opus 4.6 (best long-context + anti-injection). You need your own API key or Anthropic Pro subscription ($20/mo).

SSH Access

Power users can SSH directly into their Pod for advanced configuration, debugging, or running OpenClaw CLI commands.

$ agentputer ssh my-agent

Connecting to pod-abc123...

root@pod-abc123:~#

# Now you can run openclaw commands directly:

$ openclaw doctor

✓ All checks passed

UNDERSTAND

Security & Privacy

AgentPuter isolates each Pod in its own container. Your data stays in your Pod.
OpenClaw provides DM Pairing to control who can message your agent.

Pod Isolation

Dedicated container per user. No shared resources.

DM Pairing

OpenClaw's built-in auth. Unknown senders must be approved.

Encrypted Secrets

API keys stored encrypted, never exposed to models.

Sandbox Mode

OpenClaw Sandbox limits what skills can access.

Troubleshooting

Common issues and how to fix them.

Bot not responding?

Check Pod status: agentputer status. Verify your model API key is set and valid.

DM Pairing code not arriving?

Ensure the bot token is correct. Try: openclaw doctor via SSH.

Pod won't start?

Check logs: agentputer logs my-agent. Contact support if the issue persists.

Cost & Usage

Your costs come from two sources:

① AgentPuter (Infrastructure)

Pod hosting fee. Covers compute, storage, networking, auto-updates, and Dashboard access.

② Model API (Your subscription)

AI model usage. You bring your own API key (e.g., Anthropic Pro $20/mo).
AgentPuter provides a cost monitoring dashboard so you always know your spend.

$ agentputer usage my-agent

This month: 1,247 messages | ~$8.40 API cost

Budget limit: $50/mo | Used: 16.8%

✓ Within budget