OpenClaw Symptom-First Triage Card: Stop Guessing, Find the Right Layer

If you keep “trying fixes” without a layer model, OpenClaw will punish you with wasted time.

This page is the opposite: start from what you see, run a tiny command pack, and land on the correct fix page.

0) The 60-second command pack (gateway host)

Run on the machine/container that runs the gateway:

openclaw status --all
openclaw gateway status
openclaw models status --probe
openclaw channels status --probe
openclaw logs --limit 200 --plain

If you can only do one thing: run openclaw status --all and stop guessing.

1) Symptom cards (what you see -> what it usually is -> where to go)

A) Control UI says “unauthorized” (or keeps reconnecting)

This is usually:

• token mismatch (wrong gateway, wrong env override), or
• you opened the raw URL instead of the tokenized dashboard link.

Do this:

• Fix page: /troubleshooting/solutions/control-ui-unauthorized
• Guide (longer): /guides/control-ui-auth-and-pairing

B) Control UI / TUI says “pairing required (1008)”

This is usually:

• a new device identity that was never approved, often triggered by using a new origin (LAN/tailnet/reverse proxy).

Do this:

• Fix page: /troubleshooting/solutions/control-ui-pairing-required
• Explanation (why this happens): /guides/openclaw-pairing-explained

C) “Gateway disconnected (4008)” or “can’t connect to gateway”

This is usually:

• the gateway isn’t running,
• you’re hitting the wrong host/port,
• bind is loopback-only but you are connecting remotely,
• or auth is misaligned.

Do this:

• Start here: /guides/openclaw-cli-command-cheatsheet-and-reddit-faq
• Remote deployments: /guides/openclaw-deployment-troubleshooting

D) “It only chats” / “it won’t use tools anymore”

This is usually:

• tools.profile became narrower after an upgrade, or
• a deny rule or per-agent override is still pinning tools.

Do this:

• Fix page: /troubleshooting/solutions/openclaw-only-chats-no-tools-after-update
• Guide: /guides/openclaw-not-using-tools-after-update

E) “No output” / “nothing comes back” / “everything looks rate-limited”

This is usually:

• model auth/quota/provider failure,
• a relay/api-mode mismatch,
• or a false rate limit classification that poisons your route.

Do this:

• Guide: /guides/openclaw-no-response-and-rate-limit-troubleshooting
• Relay mismatch deep dive: /guides/openclaw-relay-and-api-proxy-troubleshooting

F) Docker/VPS “works” but Control UI is unreachable

This is usually:

• bind is still loopback,
• health checks are killing the service,
• state is not persisted (token/devices reset on restart),
• or TLS/reverse proxy expectations changed after an update.

Do this:

• Guide: /guides/openclaw-deployment-troubleshooting
• Fix page (proxy/TLS edge case): /troubleshooting/solutions/gateway-reverse-proxy-breaks-after-tls-auto-generate

G) “It stops when I close the terminal” (or after SSH disconnect)

This is usually:

• you never installed a supervisor/daemon, so the gateway is just a foreground process.

Do this:

• Fix page (Linux): /troubleshooting/solutions/linux-gateway-service-not-installed
• Guide: /guides/new-user-checklist

2) Verification (prove the fix, do not trust vibes)

After any fix:

1. rerun the 60-second command pack, and
2. do one real action:
   • dashboard refresh without loops (openclaw dashboard), or
   • send a test message on your channel.

If you cannot reproduce success twice in a row, you do not have a fix yet.