Haven't installed OpenClaw yet? Click here for one-line install commands
curl -fsSL https://openclaw.ai/install.sh | bashiwr -useb https://openclaw.ai/install.ps1 | iexcurl -fsSL https://openclaw.ai/install.cmd -o install.cmd && install.cmd && del install.cmdopenclaw doctoris OpenClaw's built-in one-stop diagnostic tool that automatically checks configuration file syntax, model authentication status, Gateway connectivity, and Node.js version compatibility[1]- Adding the
--fixflag lets doctor automatically repair most common issues, including missing configuration values, expired tokens, and corrupted workspace indexes[1] - Gateway connection failure ("Pairing Required" error) is the most common beginner issue, typically resolved simply by regenerating the Gateway Token[2]
- The root cause of agent "freezes" is usually LLM request timeouts -- increasing
timeoutSecondsor configuring a fallback model can effectively mitigate this[4]
1. openclaw doctor: Your First Line of Defense
When OpenClaw exhibits abnormal behavior, the first thing you should always do is run openclaw doctor. This command performs a comprehensive health check of the entire system and reports the status of each item with clear red/yellow/green indicators.[1]
1.1 Basic Diagnostics
openclaw doctor
This command checks the following in sequence:
- Node.js version: OpenClaw requires Node.js 22 or above
- Configuration file integrity: Syntax and required fields in
openclaw.json - Model authentication: Whether configured providers can connect successfully
- Gateway status: Whether the Gateway is running and the port is reachable
- Channel connectivity: Whether configured channels (Telegram, WhatsApp, etc.) are functioning normally
1.2 Automatic Repair
openclaw doctor --fix
With --fix added, doctor doesn't just diagnose problems -- it also attempts to repair them. Common automatic repairs include:
- Regenerating missing Gateway Tokens
- Fixing corrupted JSON5 syntax
- Rebuilding workspace indexes
- Updating expired OAuth Tokens
1.3 Generate Gateway Token
openclaw doctor --generate-gateway-token
This command is specifically used to generate or regenerate the Gateway authentication Token. When you see a "Pairing Required" error, running this single command usually resolves the issue.[2]
2. The Five Most Common Problems and Solutions
2.1 "Pairing Required" -- Gateway Connection Failure
Symptoms: You see the "Pairing Required" message in the Web UI (http://127.0.0.1:18789) or messaging channels.
Cause: Gateway Token is missing, expired, or mismatched.
Solution:
openclaw doctor --generate-gateway-token
# Then restart the Gateway
openclaw gateway restart
2.2 "LLM Request Timed Out" -- Model Request Timeout
Symptoms: The agent stops responding mid-task, and logs show a timeout error.[4]
Cause: LLM API response time exceeds the default limit, common during complex tasks or API traffic peaks.
Solution:
# Increase timeout (in seconds)
openclaw config set agents.defaults.timeoutSeconds 600
# Set a fallback model to automatically switch when the primary model times out
openclaw config set agents.defaults.model.fallbacks '["claude-sonnet-4-6"]'
2.3 Agent "Frozen" and Unresponsive
Symptoms: The agent is completely unresponsive -- no replies and no errors after sending messages.
Cause: Possible Gateway process freeze, Node process crash, or corrupted workspace state.
Solution (from least to most disruptive):
# 1. Restart the Gateway
openclaw gateway restart
# 2. If Gateway cannot restart normally, force stop then restart
openclaw gateway stop
openclaw gateway start
# 3. If still unresolved, run a full diagnostic repair
openclaw doctor --fix
2.4 Port 18789 Already in Use
Symptoms: Starting the Gateway shows the port is already in use.
Solution:
# Check what's using the port
lsof -i :18789
# Option A: Kill the occupying process
kill -9 PID
# Option B: Change the port
openclaw config set gateway.port 28789
2.5 Model Authentication Failure
Symptoms: The agent reports "Model not allowed" or "Authentication failed".[4]
Solution:
# Check model status
openclaw models status
# Reconfigure API key
openclaw models auth setup-token --provider anthropic
# Or redo the OAuth flow
openclaw models auth login --provider openai
3. Restart and Reset Procedures
3.1 Normal Restart (Preserves All Data)
openclaw gateway restart
This is the safest restart method. The Gateway writes the current state to persistent storage before restarting. All workspaces, memory, and settings are preserved.
3.2 Full Stop and Start
openclaw gateway stop
# Wait a few seconds
openclaw gateway start
Use this when gateway restart doesn't work. The stop command ensures all related processes are fully terminated.
3.3 Daemon Mode Restart (systemd)
If you installed the systemd service using openclaw onboard --install-daemon:
systemctl --user restart openclaw-gateway
3.4 Clear Context (Soft Reset)
If the agent's conversation context becomes confused, you can clear just the context without restarting the entire system:
openclaw tui
This opens a fresh interactive conversation interface without affecting long-term memory in the workspace.
4. Log Analysis and Advanced Debugging
4.1 View Real-Time Logs
openclaw logs --follow
This command outputs Gateway runtime logs in real time, similar to tail -f. You can see the processing of each request, model call latency, and error messages.[7]
4.2 Common Log Message Reference
| Log Message | Meaning | Recommended Action |
|---|---|---|
Gateway started on :18789 | Normal startup | No action needed |
Pairing required | Token verification failed | Run doctor --generate-gateway-token |
LLM request timed out | Model response timeout | Increase timeoutSeconds or set a fallback |
Model not allowed | Model not in the allowlist | Check allowlist settings |
Agent bootstrap pending | Agent is initializing | Wait for completion, usually takes 10-30 seconds |
WebSocket connection closed | WebSocket disconnected | Check network connection or reverse proxy settings |
5. Preventive Maintenance Recommendations
Rather than reactive troubleshooting, proactive prevention is better. Here is a maintenance checklist based on our hands-on experience:[5][6]
- Run
openclaw doctoronce a week: Even when the system is running normally, regular checks can catch potential issues early - Keep OpenClaw updated: Run
npm update -g openclawto get the latest security patches and feature improvements - Configure fallback models: Never rely on a single model provider
- Monitor Gateway memory usage: Long-running Gateways may develop memory leaks; restarting weekly is recommended
- Back up the
~/.openclaw/directory: Contains all your settings, authentication, and workspace data
Conclusion
openclaw doctor with the --fix flag can automatically resolve over 80% of common issues.[1] For more complex problems, the log analysis and systematic troubleshooting procedures provided in this article can help you quickly identify the root cause.
If you're just getting started with OpenClaw, we recommend first reading our Architecture Analysis & Hands-On Deployment Guide to build a complete understanding of the system; if you need to adjust settings, refer to the Configuration Complete Guide.



