OpenClaw Settings and Monitoring Dashboard

A static control page for the local OpenClaw agent running on A8MAX. It documents the working Telegram setup, the OpenAI Codex primary model, the Gemma fallback, and the commands used to verify health.

Verified: 2026-05-11 20:48 KST Gateway: ws://127.0.0.1:18789 Bot: @swc_openclawbot Model: gpt-5.5

Live Operations View

Use this as the first screen when checking whether the local agent is available from Telegram.

Gateway

Reachable

Local dashboard at http://127.0.0.1:18789/.

Configured

Direct messages are paired to account 7400509226.

Model Router

gpt-5.5

Fallback: ollama/gemma4:latest.

Tool Profile

Coding

Shell, filesystem editing, gateway, browser, and web tools are allowed.

Gateway Probe

Optional browser-side test. GitHub Pages may block local WebSocket checks in some browsers.

WebSocket URL

Probe state Not tested

Last checked Never

Session Health

Most failures should be checked in this order before changing configuration.

Gateway process is running openclaw status reports reachable loopback gateway and active sessions. Telegram delivery is single-path The custom delivery bridge is stopped, so duplicate replies should not reappear. Fallback model is installed Gemma4 is available through Ollama if the OpenAI subscription quota is limited.

Settings Snapshot

Redacted operational settings from the local OpenClaw host. Tokens are intentionally not published.

Area	Value	Notes
Config file	`~/.openclaw/openclaw.json`	Contains channel, model, tool, and plugin configuration.
Gateway	`ws://127.0.0.1:18789`	Local loopback only. Tailscale exposure is off.
Dashboard	`http://127.0.0.1:18789/`	If the browser reports an expired device signature, reload the dashboard from a fresh OpenClaw URL.
Telegram	`@swc_openclawbot`	Enabled with pairing policy and progress streaming.
Primary model	`openai-codex/gpt-5.5`	Uses the Codex login profile rather than a pay-as-you-go API key.
Fallback model	`ollama/gemma4:latest`	Used when primary routing is unavailable or quota constrained.
Tool profile	`coding`	Restores shell, filesystem editing, browser, gateway, and web fetch/search tools.
Memory	`memory-lancedb`	Enabled. Status reports memory as not checked in fast mode.

Monitoring Commands

Run these on the host PC. Copy buttons store only the command text, not credentials.

Fast Status

Overall agent, channels, sessions

openclaw status

Deep Probe

Channel and environment checks

openclaw status --deep

Gateway Health

Local gateway and Telegram

openclaw gateway health

Eligible Skills

Ready-to-use skill list

openclaw skills list --eligible

Model Status

Primary model route

openclaw models status --plain

Fallbacks

Backup model chain

openclaw models fallbacks list

Live Logs

Debug message flow

openclaw logs --follow

Dashboard URL

Fresh local dashboard link

openclaw dashboard

Restart Gateway

Use after config changes

openclaw gateway run

Ready Skills

After installing and authenticating GitHub CLI, the GitHub skills are ready too.

12 skills shown

browser-automation ready Controls web pages with the OpenClaw browser tool for multi-step flows.

gh-issues ready Fetches issues, delegates fixes, opens pull requests, and watches reviews.

github ready Uses gh for issues, PRs, CI logs, reviews, releases, and API queries.

healthcheck ready Audits and hardens hosts running OpenClaw.

node-connect ready Diagnoses node pairing, QR/setup codes, routes, and auth failures.

skill-creator ready Creates, edits, improves, reviews, and audits AgentSkills.

taskflow ready Coordinates multi-step detached tasks as durable jobs.

taskflow-inbox-triage ready Example pattern for inbox triage, intent routing, replies, and summaries.

tmux ready Remote-controls tmux sessions for interactive CLIs.

video-frames ready Extracts frames or short clips from videos using ffmpeg.

voice-call ready Starts voice calls through the OpenClaw voice-call plugin.

weather ready Gets current weather, rain, temperature, and forecasts.

Message Flow

The Telegram bot is the remote control surface. The PC remains the execution host.

Remote Conversation Model

You can talk to the PC from a walk as long as the host stays awake, the OpenClaw gateway is running, the Telegram bot is reachable, and the network connection is alive. The phone is only the message client; shell commands and file edits still execute on the PC.

Do not publish gateway tokens, Telegram bot tokens, or OAuth files. This page records only redacted settings and operational commands.

Troubleshooting

Known symptoms from this setup and the shortest useful checks.

No reply Check the gateway and Telegram channel first. Run openclaw gateway health, then openclaw logs --follow. If the gateway is not running, start it again with openclaw gateway run.

Double reply Look for duplicate Telegram senders. The built-in Telegram provider should be the only delivery path. Stop any custom bridge process before retesting.

Dashboard auth Expired device signatures are local dashboard auth issues. Use openclaw dashboard to generate a fresh URL. This does not necessarily mean Telegram is broken.

Missing tools Verify the tool profile is still coding. If Telegram reports no shell, filesystem, browser, gateway, or config tools, inspect tools.profile and restart the gateway after changing it.

Quota Use Gemma4 when Codex quota is unavailable. The configured fallback is ollama/gemma4:latest. Confirm it with openclaw models fallbacks list.