Aleph Cloud Self-Deployment

Teach an AI agent to deploy itself onto Aleph Cloud decentralized infrastructure — from provisioning a VM to configuring its own runtime, autonomously.

Overview

This skill gives an AI agent everything it needs to:

1. Provision its own VM on the Aleph Cloud network
2. Install its agent framework (OpenClaw) on the new instance
3. Configure its own credentials, gateway, and identity
4. Transfer its knowledge and skills to the new environment
5. Become fully operational on decentralized compute — without human intervention

Prerequisites

• aleph-client CLI installed (pip install aleph-client)
• A funded Aleph Cloud account (credits on BASE/ETH/AVAX/SOL, or $ALEPH tokens for holding)
• An SSH keypair for VM access
• AI provider API key (Anthropic, OpenAI, etc.) to share with spawned agents

Install aleph-client

pip install aleph-client --break-system-packages

On Ubuntu 24+ you need --break-system-packages or use a venv.

Check Version

aleph --version

Step 1: Account Setup

Import or Create a Private Key

# Import existing private key
aleph account create --private-key \x3CYOUR_HEX_PRIVATE_KEY> --chain BASE --active

# Or generate a new one
aleph account create --chain BASE --active

Keys are stored at ~/.aleph-im/private-keys/. Always chmod 600 key files.

Check Balance

aleph account balance

You need credits or $ALEPH tokens. Credits are the simplest (pay-as-you-go). Buy credits at https://account.aleph.im/ or hold $ALEPH tokens.

Pricing Reference

aleph pricing instance

Disk size can be overridden with --rootfs-size \x3CMiB>. The tier sets the minimum; you can allocate more (e.g., 40GB on a Tier 1). Extra disk costs ~4,416 credits/GiB/day.

Step 2: Generate SSH Keypair

ssh-keygen -t ed25519 -f ~/.ssh/aleph_agent -N "" -C "agent@aleph"
chmod 600 ~/.ssh/aleph_agent

Step 3: Find a Compute Resource Node (CRN)

List Active CRNs (Top 10 by Score)

aleph node compute --active --json 2>/dev/null | python3 -c "
import sys, json
nodes = json.load(sys.stdin)
scored = []
for n in nodes:
    score = n.get('score', 0) or 0
    name = n.get('name', 'unknown')
    node_hash = n.get('hash', '')
    url = n.get('address', '')
    if score > 0.5:  # Score is 0.0–1.0 (not percentage)
        scored.append((score, name, node_hash, url))
scored.sort(reverse=True)
for s, name, h, url in scored[:10]:
    print(f'{s*100:.1f}% | {name} | {h} | {url}')
"

Get Best CRN Hash (for scripts)

CRN_HASH=$(aleph node compute --active --json 2>/dev/null | python3 -c "
import sys, json
nodes = json.load(sys.stdin)
best = max(
    (n for n in nodes if (n.get('score', 0) or 0) > 0.5),
    key=lambda n: n.get('score', 0),
    default=None
)
print(best['hash'] if best else '')
")
echo "Best CRN: $CRN_HASH"

Step 4: Create the VM Instance

Known Rootfs Hashes

Create Instance (Non-Interactive)

The aleph instance create command has an interactive TUI for CRN selection. To bypass it completely, use --crn-hash + --skip-volume:

aleph instance create \
  --name "agent-v1" \
  --compute-units 1 \
  --rootfs 5330dcefe1857bcd97b7b7f24d1420a7d46232d53f27be280c8a7071d88bd84e \
  --rootfs-size 40960 \
  --payment-type credit \
  --payment-chain BASE \
  --ssh-pubkey-file ~/.ssh/aleph_agent.pub \
  --crn-hash "$CRN_HASH" \
  --crn-auto-tac \
  --skip-volume \
  --verbose

Critical flags:

• --crn-hash \x3CHASH>: Bypasses the interactive CRN picker TUI entirely
• --crn-auto-tac: Auto-accepts CRN Terms & Conditions
• --skip-volume: Skips the "add extra volumes?" interactive prompt
• --rootfs: OS image hash (omit for Ubuntu 22 default)
• --rootfs-size: Disk size in MiB (40960 = 40GB; overrides tier default)

If Interactive TUI Still Appears

Some versions of aleph-client may still show prompts. Use pexpect:

#!/usr/bin/env python3
"""Automate aleph instance create when CLI flags don't fully bypass TUI."""
import pexpect
import sys
import re

CRN_HASH = sys.argv[1] if len(sys.argv) > 1 else ""
NAME = sys.argv[2] if len(sys.argv) > 2 else f"agent-{int(__import__('time').time())}"

cmd = (
    f'aleph instance create '
    f'--name "{NAME}" '
    f'--compute-units 1 '
    f'--rootfs 5330dcefe1857bcd97b7b7f24d1420a7d46232d53f27be280c8a7071d88bd84e '
    f'--rootfs-size 40960 '
    f'--payment-type credit '
    f'--payment-chain BASE '
    f'--ssh-pubkey-file ~/.ssh/aleph_agent.pub '
    f'--crn-auto-tac '
    f'--skip-volume '
    f'--verbose'
)
if CRN_HASH:
    cmd += f' --crn-hash {CRN_HASH}'

child = pexpect.spawn('/bin/bash', ['-c', cmd], timeout=180, encoding='utf-8')
child.logfile_read = sys.stdout

while True:
    try:
        idx = child.expect([
            r'(?i)rootfs.*type',         # Rootfs type prompt
            r'(?i)disk.*size',           # Disk size prompt
            r'(?i)add.*volume',          # Add volume prompt
            r'(?i)confirm|deploy|proceed', # Confirmation prompt
            r'(?i)select.*crn|choose.*node', # CRN selection
            pexpect.EOF,
            pexpect.TIMEOUT,
        ], timeout=120)

        if idx == 0:
            child.sendline('ubuntu24')
        elif idx == 1:
            child.sendline('40960')
        elif idx == 2:
            child.sendline('n')
        elif idx == 3:
            child.sendline('y')
        elif idx == 4:
            child.sendline('')  # Press Enter for top/default choice
        elif idx == 5:
            break  # EOF — done
        elif idx == 6:
            break  # Timeout — done or stuck
    except Exception as e:
        print(f"Error: {e}", file=sys.stderr)
        break

print("\
" + child.before if child.before else "")
child.close()
sys.exit(child.exitstatus or 0)

Install pexpect: pip install pexpect --break-system-packages

Usage: python3 create_instance.py \x3CCRN_HASH> \x3CINSTANCE_NAME>

Parse Instance Hash After Creation

The CLI outputs the instance details. Extract the hash:

# List instances and get the newest one
aleph instance list 2>&1

The output shows a table with instance hash, SSH connection details, and mapped ports. Look for:

ssh root@\x3CHOST_IP> -p \x3CPORT> -i \x3Cssh-private-key-file>

Programmatic SSH Details

# Parse connection info for the most recent instance
aleph instance list 2>&1 | grep -oP 'ssh root@\S+ -p \d+'

Wait and Connect

# Wait for VM to boot (30-60 seconds)
sleep 45

# Connect
ssh -o StrictHostKeyChecking=no root@\x3CHOST_IP> -p \x3CMAPPED_PORT> -i ~/.ssh/aleph_agent

Step 5: Install OpenClaw on the New VM

SSH into the VM, then:

# Install Node.js 22
curl -fsSL https://deb.nodesource.com/setup_22.x | bash -
DEBIAN_FRONTEND=noninteractive apt-get install -y nodejs git

# Install OpenClaw
npm install -g openclaw

# Create workspace
mkdir -p /root/openclaw/memory /root/openclaw/skills

# Verify
node --version    # Should show v22.x
openclaw --version

Use DEBIAN_FRONTEND=noninteractive to prevent dpkg prompts in non-TTY sessions.

Step 6: Configure Agent Auth & Gateway

Set Up Anthropic API Key

mkdir -p /root/.openclaw/agents/main/agent /root/.openclaw/agents/main/sessions

cat > /root/.openclaw/agents/main/agent/auth-profiles.json \x3C\x3C 'EOF'
{
  "version": 1,
  "profiles": {
    "anthropic:default": {
      "type": "token",
      "provider": "anthropic",
      "token": "\x3CYOUR_ANTHROPIC_API_KEY>"
    }
  },
  "lastGood": {
    "anthropic": "anthropic:default"
  }
}
EOF
chmod 600 /root/.openclaw/agents/main/agent/auth-profiles.json

Critical: The file MUST be named auth-profiles.json (not auth.json). The gateway reads auth-profiles.json — using the wrong filename silently fails with "No API key found for provider".

Configure OpenClaw

GATEWAY_TOKEN=$(openssl rand -hex 24)
echo "Gateway token: $GATEWAY_TOKEN"  # Save this!

cat > /root/.openclaw/openclaw.json \x3C\x3C EOF
{
  "auth": {
    "profiles": {
      "anthropic:default": {
        "provider": "anthropic",
        "mode": "token"
      }
    }
  },
  "agents": {
    "defaults": {
      "model": {
        "primary": "anthropic/claude-sonnet-4-20250514"
      },
      "workspace": "/root/openclaw"
    }
  },
  "gateway": {
    "port": 18789,
    "mode": "local",
    "bind": "loopback",
    "auth": {
      "mode": "token",
      "token": "$GATEWAY_TOKEN"
    }
  }
}
EOF

# Harden permissions
chmod 700 /root/.openclaw
chmod 600 /root/.openclaw/openclaw.json

Start Gateway

openclaw gateway install --force
openclaw gateway start

# Wait for startup
sleep 3

# Verify — MUST pass the token
openclaw gateway status --token "$GATEWAY_TOKEN"
# Should show: Runtime: running ... RPC probe: ok

If RPC probe: failed after gateway start, run openclaw gateway install --force then openclaw gateway restart. The systemd service file must have the token synced.

Test Agent

OPENCLAW_GATEWAY_TOKEN="$GATEWAY_TOKEN" openclaw agent --agent main -m "Hello! Who are you? Reply in one sentence."

If the gateway route fails, it falls back to embedded mode (still works, just slower on first call).

Step 7: Copy Identity & Skills

From the parent machine, copy workspace files to the new VM:

HOST="\x3CHOST_IP>"
PORT="\x3CMAPPED_PORT>"
KEY="$HOME/.ssh/aleph_agent"
SCP="scp -i $KEY -P $PORT -o StrictHostKeyChecking=no"

# Core identity files
for f in SOUL.md AGENTS.md IDENTITY.md USER.md TOOLS.md HEARTBEAT.md; do
  [ -f "/root/openclaw/$f" ] && $SCP /root/openclaw/$f root@$HOST:/root/openclaw/$f
done

# Memory (optional — fresh start or shared context)
[ -f "/root/openclaw/MEMORY.md" ] && $SCP /root/openclaw/MEMORY.md root@$HOST:/root/openclaw/MEMORY.md

# Skills directory (recursive)
scp -r -i $KEY -P $PORT -o StrictHostKeyChecking=no /root/openclaw/skills/ root@$HOST:/root/openclaw/skills/

Step 8: Enable Self-Deployment

For the new agent to deploy new instances, it needs:

1. aleph-client CLI on its VM
2. Aleph private key (or delegated key) to create instances
3. This skill in its skills directory
4. Sufficient credits/tokens in the account

Install aleph-client on the Child VM

ssh -i $KEY -p $PORT root@$HOST \x3C\x3C 'REMOTE'
apt-get install -y python3-pip
pip install aleph-client --break-system-packages

# Verify
aleph --version
REMOTE

Transfer Aleph Private Key

# SENSITIVE — never log this key
$SCP /root/.aleph-im/private-keys/your-key.key root@$HOST:/root/.aleph-im/private-keys/your-key.key
ssh -i $KEY -p $PORT root@$HOST "chmod 700 /root/.aleph-im/private-keys && chmod 600 /root/.aleph-im/private-keys/*.key"

# Activate on the child
ssh -i $KEY -p $PORT root@$HOST "aleph account create --private-key \$(cat /root/.aleph-im/private-keys/your-key.key) --chain BASE --active"

Verify Self-Deployment Capability

ssh -i $KEY -p $PORT root@$HOST \x3C\x3C 'REMOTE'
echo "=== Checking deployment readiness ==="
echo -n "aleph-client: "; aleph --version
echo -n "Balance: "; aleph account balance 2>&1 | head -3
echo -n "SSH key: "; ls ~/.ssh/aleph_agent.pub 2>/dev/null && echo "OK" || echo "MISSING — generate with: ssh-keygen -t ed25519 -f ~/.ssh/aleph_agent -N ''"
echo -n "This skill: "; ls ~/openclaw/skills/aleph-vm-deployment/SKILL.md 2>/dev/null && echo "OK" || echo "MISSING"
echo -n "OpenClaw: "; openclaw --version
echo "=== Ready to deploy ==="
REMOTE

Now the child instance can read this SKILL.md and create new instances — recursively.

Step 9: Manage Instances

List All Instances

aleph instance list

Delete an Instance

aleph instance delete \x3CINSTANCE_HASH>

Check Account Balance (Monitor Burn Rate)

aleph account balance

Deployment Script (All-in-One)

This script automates the full cycle: find CRN → create VM → wait → install OpenClaw → configure → verify.

#!/bin/bash
# deploy.sh — Create a new Aleph Cloud agent clone
# Usage: ./deploy.sh [name] [compute-units] [anthropic-key]
set -euo pipefail

AGENT_NAME="${1:-agent-$(date +%s)}"
COMPUTE_UNITS="${2:-1}"
ANTHROPIC_KEY="${3:-$ANTHROPIC_API_KEY}"
SSH_KEY="$HOME/.ssh/aleph_agent"
ROOTFS="5330dcefe1857bcd97b7b7f24d1420a7d46232d53f27be280c8a7071d88bd84e"
ROOTFS_SIZE=40960

if [ -z "$ANTHROPIC_KEY" ]; then
    echo "ERROR: Set ANTHROPIC_API_KEY env var or pass as 3rd arg"
    exit 1
fi

# Generate SSH key if needed
if [ ! -f "$SSH_KEY" ]; then
    echo "=== Generating SSH keypair ==="
    ssh-keygen -t ed25519 -f "$SSH_KEY" -N "" -C "$AGENT_NAME@aleph"
fi

# Find best CRN
echo "=== Finding best CRN ==="
CRN_HASH=$(aleph node compute --active --json 2>/dev/null | python3 -c "
import sys, json
nodes = json.load(sys.stdin)
best = max(
    (n for n in nodes if (n.get('score', 0) or 0) > 0.5),
    key=lambda n: n.get('score', 0),
    default=None
)
print(best['hash'] if best else '')
")

if [ -z "$CRN_HASH" ]; then
    echo "ERROR: No suitable CRN found"
    exit 1
fi
echo "Selected CRN: $CRN_HASH"

# Create instance
echo "=== Creating instance: $AGENT_NAME ==="
aleph instance create \
    --name "$AGENT_NAME" \
    --compute-units "$COMPUTE_UNITS" \
    --rootfs "$ROOTFS" \
    --rootfs-size "$ROOTFS_SIZE" \
    --payment-type credit \
    --payment-chain BASE \
    --ssh-pubkey-file "${SSH_KEY}.pub" \
    --crn-hash "$CRN_HASH" \
    --crn-auto-tac \
    --skip-volume \
    --verbose

# Wait for boot
echo "=== Waiting 45s for VM to boot ==="
sleep 45

# Get connection details
echo "=== Connection details ==="
aleph instance list 2>&1 | tail -30

# Parse SSH connection (user must extract HOST and PORT from output above)
echo ""
echo "=== $AGENT_NAME created ==="
echo ""
echo "Next steps:"
echo "  1. Find HOST and PORT from the instance list above"
echo "  2. SSH in: ssh root@\x3CHOST> -p \x3CPORT> -i $SSH_KEY"
echo "  3. Run the setup script:"
echo "     ssh root@\x3CHOST> -p \x3CPORT> -i $SSH_KEY 'bash -s' \x3C setup-agent.sh"

Post-Creation Setup Script

Save this as setup-agent.sh and pipe it via SSH:

#!/bin/bash
# setup-agent.sh — Run on a fresh Aleph VM after creation
# Usage: ssh root@HOST -p PORT -i KEY 'bash -s' \x3C setup-agent.sh
set -e

ANTHROPIC_KEY="${1:-}"

echo "=== Installing Node.js 22 ==="
curl -fsSL https://deb.nodesource.com/setup_22.x | bash -
DEBIAN_FRONTEND=noninteractive apt-get install -y nodejs git python3-pip

echo "=== Installing OpenClaw ==="
npm install -g openclaw

echo "=== Installing aleph-client ==="
pip install aleph-client --break-system-packages

echo "=== Creating workspace ==="
mkdir -p /root/openclaw/{memory,skills}
mkdir -p /root/.openclaw/agents/main/{agent,sessions}

echo "=== Configuring gateway ==="
GATEWAY_TOKEN=$(openssl rand -hex 24)

cat > /root/.openclaw/openclaw.json \x3C\x3C CONF
{
  "auth": {
    "profiles": {
      "anthropic:default": {
        "provider": "anthropic",
        "mode": "token"
      }
    }
  },
  "agents": {
    "defaults": {
      "model": {
        "primary": "anthropic/claude-sonnet-4-20250514"
      },
      "workspace": "/root/openclaw"
    }
  },
  "gateway": {
    "port": 18789,
    "mode": "local",
    "bind": "loopback",
    "auth": {
      "mode": "token",
      "token": "$GATEWAY_TOKEN"
    }
  }
}
CONF

chmod 700 /root/.openclaw
chmod 600 /root/.openclaw/openclaw.json

echo "=== Starting gateway ==="
openclaw gateway install --force
openclaw gateway start
sleep 3

echo ""
echo "=== SETUP COMPLETE ==="
echo "Gateway token: $GATEWAY_TOKEN"
echo "OpenClaw: $(openclaw --version)"
echo "Node: $(node --version)"
echo ""
echo "Remaining: copy auth-profiles.json, identity files, and aleph key"

Security Considerations

• Private keys: Always chmod 600. Never commit to git. Never log or echo.
• API keys: Store in auth-profiles.json with chmod 600, not in env vars or config files.
• Gateway tokens: Generate unique per VM (openssl rand -hex 24). Sync with openclaw gateway install --force.
• SSH: Use ed25519 keys only. Disable password auth on production VMs.
• Instance limits: Set a maximum instance count before deploying. Monitor credit balance and set budget alerts to prevent cost overruns.
• Human approval: Always require human approval before provisioning new instances in production.
• Network isolation: Gateway binds to loopback by default. Only expose via Tailscale, never raw public IP.
• Key isolation: All instances sharing one Aleph key is a security risk. Use delegated keys for production.

Cost Planning

Extra disk beyond tier default: ~4,416 credits/GiB/day.

Check current pricing: aleph pricing instance

Troubleshooting

"No API key found for provider" Error

The gateway reads auth-profiles.json, NOT auth.json. Ensure:

• File is at ~/.openclaw/agents/main/agent/auth-profiles.json
• Format matches the template in Step 6 (with "version": 1 and "profiles" key)
• Run openclaw gateway restart after any auth changes

RPC Probe Failed

openclaw gateway install --force
openclaw gateway restart
sleep 3
openclaw gateway status --token "$GATEWAY_TOKEN"

The systemd service file caches the token — install --force re-syncs it.

Interactive TUI Blocks Automation

Use --crn-hash \x3CHASH> + --skip-volume + --crn-auto-tac. If still interactive, use the pexpect script from Step 4.

VM Not Reachable After Creation

• Wait 45-60 seconds (boot takes time on decentralized infra)
• Check aleph instance list for the mapped SSH port
• Verify the CRN is healthy: aleph node compute --crn-hash \x3CHASH> --json
• Try IPv6 if IPv4 mapping isn't working

Credit Balance Dropping Too Fast

• aleph account balance — check remaining credits
• aleph instance list — count running instances
• aleph instance delete \x3CHASH> — kill unused instances
• Each running instance burns credits continuously, even idle

Config Validation Errors ("Unrecognized key")

Only use config keys that OpenClaw recognizes. Known valid top-level keys: auth, agents, gateway, models, tools, bindings, messages, commands, hooks, channels, skills, plugins, wizard, meta

Run openclaw doctor --fix to auto-remove invalid keys.

dpkg Prompts Block SSH Scripts

Use DEBIAN_FRONTEND=noninteractive before apt commands in non-TTY sessions.