Skip to main content

​
Agent Operator Guide (OpenClaw)

This guide is for operators who want the full stack:
  • b1e55ed (signals + brain + dashboard)
  • OpenClaw (24/7 AI operator)
  • Telegram alerts + natural-language queries
  • Heartbeat health checks
If you want no AI dependency, start here instead:

​
What you get

Everything in the standalone deployment, plus:
  • AI monitors 24/7: watches health, events, and anomalies
  • Telegram alerts: sends actionable notifications (failures, thresholds, noteworthy shifts)
  • Natural-language ops: ask questions like “what’s HYPE doing?” and get a synthesized answer
  • Proactive analysis: agent can summarize what changed since last check
  • Heartbeat checks: a small, explicit checklist that runs periodically

​
Architecture

                (your private VPS)

        +-----------------------------+
        |          b1e55ed             |
        |  - producers + brain loop    |
        |  - REST API (:5050)          |
        |  - dashboard (:5051)         |
        +--------------^--------------+
                       |
                       | localhost / tailscale
                       v
        +-----------------------------+
        |          OpenClaw           |
        |  - AI layer + heartbeats    |
        |  - alert routing + tools    |
        +--------------^--------------+
                       |
                       v
                  Telegram (Bot)

Optional: Oracle endpoint exposed so agents (or other systems) can check producer provenance.

​
Prerequisites

​
Same as standalone

  • VPS: 2 vCPU / 2 GB RAM minimum (4 GB recommended)
  • Python 3.11+ (installer + uv manage this)
  • GitHub access (installer pulls from GitHub)

​
Additional

⚠️ Common failure point #1: no AI key / wrong AI key Symptoms: OpenClaw starts but can’t run tools, or errors on model calls. Fix: re-run openclaw setup and ensure the key is stored (see Step 4).

​
Step 1: Install b1e55ed

Do the standalone install first: 
# (Recommended) run as a dedicated operator user
sudo useradd -m -s /bin/bash b1e55ed
sudo su - b1e55ed

curl -sSf https://raw.githubusercontent.com/P-U-C/b1e55ed/main/install.sh | bash
export PATH="$HOME/.local/bin:$PATH"

b1e55ed wizard
At the end of the wizard you should be able to run b1e55ed start. Do not run it manually - Step 5 installs it as a systemd service automatically. Just confirm the wizard completes cleanly.

​
Step 2: Install OpenClaw

Install Node.js first if you don’t have it (Ubuntu): 
curl -fsSL https://deb.nodesource.com/setup_22.x | sudo -E bash -
sudo apt-get install -y nodejs
node --version
npm --version
Install OpenClaw: 
sudo npm install -g openclaw
openclaw --version
Run setup (stores provider keys and creates a default workspace): 
openclaw setup
During setup, when prompted for an Anthropic key:

​
Step 3: Connect Telegram

​
3.1 Create a Telegram bot

  1. In Telegram, open @BotFather
  2. Send /newbot
  3. Choose a name and username
  4. BotFather will return a bot token (a long string)

​
3.2 Store the token in OpenClaw

On your VPS: 
# Replace TOKEN with the exact token BotFather gave you.
# Example format: 1234567890:AA... (keep it secret)
openclaw config set telegram.token "TOKEN"

​
3.3 Test the bot

  1. Open your bot in Telegram
  2. Tap Start (or send /start)
If OpenClaw is running, you should see it acknowledge messages (depending on your configured skills/heartbeats). ⚠️ Common failure point #2: Telegram bot doesn’t respond Checklist: 
# Confirm token was set
openclaw config get telegram.token

# Check OpenClaw logs (if running under systemd, see later section)
openclaw gateway status || true
Also confirm you are messaging the correct bot (the one you created with BotFather).

​
Step 4: Configure the AI (SOUL.md + USER.md)

OpenClaw uses a workspace directory containing operator instructions. Two files matter most:
  • SOUL.md: the assistant’s personality + operating principles
  • USER.md: your preferences and what “good” looks like
The setup script in Step 5 writes these for you from your answers to the prompts. You do not need to create them manually. If you want to customize after setup, edit them at ~/.openclaw/workspace/ — your own copy with your real name, Telegram handle, timezone, and node ID already filled in.

​
Step 5: Connect b1e55ed to OpenClaw

The simplest integration pattern:
  • b1e55ed runs on the VPS (API on 127.0.0.1:5050)
  • OpenClaw heartbeats call:
    • b1e55ed health --json
    • curl http://127.0.0.1:5050/api/v1/health
    • b1e55ed status

​
5.0 Run the operator setup scripts

Two separate scripts, two separate concerns: scripts/setup-connected.sh — installs b1e55ed + OpenClaw + Telegram, and registers b1e55ed as a systemd service. Run this first if you haven’t already (it’s the full connected install script). scripts/setup-openclaw.sh — configures the OpenClaw workspace (USER.md, CRITICAL.md, queue-drain cron). Run this after.
Do Step 3 (Telegram bot) before running setup-openclaw.sh. The script will ask for your bot token.
export GH_TOKEN="ghp_xxx"   # required for queue automation

# If starting fresh (installs everything + systemd):
bash scripts/setup-connected.sh

# OpenClaw workspace only (USER.md, CRITICAL.md, cron):
bash scripts/setup-openclaw.sh
setup-openclaw.sh prompts for: name, Telegram handle, timezone, GitHub username, bot token — then writes your workspace files with real values (no placeholders). Verify the engine is running: 
sudo systemctl status b1e55ed     # engine running?
sudo journalctl -u b1e55ed -f     # live logs
openclaw cron list                 # queue drain active?

​
5.1 Ensure b1e55ed is reachable locally

On the VPS (give it ~10s to start): 
curl -s http://127.0.0.1:5050/api/v1/health

​
5.2 Create a HEARTBEAT.md

In your OpenClaw workspace, create HEARTBEAT.md with explicit checks. Minimal, copy/paste-ready example: 
Every heartbeat:

1) b1e55ed health (cron-safe)
   - run: b1e55ed health --json
   - alert if status != OK

2) API reachable
   - run: curl -s http://127.0.0.1:5050/api/v1/health
   - alert if not HTTP 200 or body is not healthy

3) Brain freshness
   - run: b1e55ed status
   - alert if last brain cycle is stale (>10 minutes)
Then run a manual check to validate your environment (optional but recommended): 
b1e55ed health --json
b1e55ed status
⚠️ Common failure point #3: OpenClaw can’t find the b1e55ed command Fix: ensure the service environment PATH includes ~/.local/bin, or use an absolute path in your heartbeat commands.
Tailscale lets you access the dashboard/API privately without exposing ports to the public internet.

​
VPS install

curl -fsSL https://tailscale.com/install.sh | sh
sudo tailscale up
After tailscale up, it will print a login URL. Open it, authenticate, and approve the device.

​
Laptop (macOS) install

Now you can bind b1e55ed to localhost and still reach it over your tailnet by using the VPS Tailscale IP and a reverse proxy or tailscale serve (optional). For production hardening, see docs/deployment.md.

​
Step 7: Oracle setup (end-to-end)

If you want contributor registrations and agent provenance checks to work end-to-end, ensure:
  • API is running
  • Oracle endpoint is reachable wherever your agent runs
Oracle docs:

​
Daily operations

​
What alerts look like

Typical Telegram alerts you should configure for:
  • services restarting (API/brain/OpenClaw)
  • b1e55ed health returning DEGRADED/ERROR
  • brain staleness (no successful cycles within your threshold)

​
Example natural-language queries

  • “what’s BTC doing today?”
  • “summarize last 2 hours of signals”
  • “is the engine healthy?”
Under the hood, OpenClaw should answer by calling:
  • b1e55ed status
  • b1e55ed positions --json
  • API health endpoints

​
How heartbeats work

Heartbeats are deliberate, small checklists. You define them in HEARTBEAT.md. The agent executes them on a schedule and alerts on violations.

​
Customizing alerts

Edit HEARTBEAT.md to change thresholds and escalation. Practical knobs:
  • “brain stale” threshold (e.g., 10 minutes vs 30 minutes)
  • which endpoints matter (API, oracle, dashboard)
  • noise filtering (only alert on consecutive failures)
Keep it explicit and small so it stays reliable.

​
Running as a service (systemd)

This section runs two services under systemd:
  • b1e55ed.service — runs b1e55ed daemon (manages API, dashboard, brain, resolver internally)
  • openclaw-gateway.service — runs the OpenClaw AI layer

​
b1e55ed service

b1e55ed daemon manages all subsystems in a single process — no separate API/brain units, no crontab. See: Standalone Operator Guide → Running as a service

​
OpenClaw gateway service

This assumes:
  • OpenClaw is installed globally (/usr/bin/openclaw should exist)
  • You want it running as the b1e55ed user
sudo tee /etc/systemd/system/openclaw-gateway.service >/dev/null <<'EOF'
[Unit]
Description=OpenClaw gateway
After=network.target

[Service]
Type=simple
User=b1e55ed
WorkingDirectory=/home/b1e55ed
Environment=PATH=/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/home/b1e55ed/.local/bin
ExecStart=/usr/bin/openclaw gateway start
ExecStop=/usr/bin/openclaw gateway stop
Restart=always
RestartSec=5

[Install]
WantedBy=multi-user.target
EOF
Enable + start: 
sudo systemctl daemon-reload
sudo systemctl enable --now openclaw-gateway.service
systemctl status openclaw-gateway.service --no-pager
If your OpenClaw install path differs, locate it: 
command -v openclaw

​
Troubleshooting

​
OpenClaw installed but openclaw not found

npm -g bin
command -v openclaw
sudo npm install -g openclaw

​
Telegram connection issues

  • Re-check token with openclaw config get telegram.token
  • Ensure the bot was started in Telegram (/start)

​
b1e55ed API not reachable from OpenClaw

curl -v http://127.0.0.1:5050/api/v1/health
systemctl status b1e55ed.service --no-pager || true
journalctl -u b1e55ed -n 200 --no-pager || true
tail -f ~/.b1e55ed/logs/api.log
For deeper b1e55ed ops docs, see the repo docs index in README.md.