Performance Troubleshooting

When Your Instance Feels Slow

If your OpenClaw instance is responding slowly, taking a long time to generate answers, or becoming unresponsive under load, this page helps identify the cause and suggests fixes.

Understand Where Time Is Spent

Response time in an OpenClaw instance depends on several factors:

Factor	Typical Impact	What Affects It
LLM response time	2-30 seconds	Model size, prompt length, provider load
Container startup	Up to 60 seconds	Only after restart or rebuild
Skill/plugin loading	1-5 seconds	Number and complexity of installed skills
Network latency	Under 100ms	Your location relative to the Falkenstein data center

The LLM provider is almost always the biggest factor in response time. If your instance "feels slow," check whether the delay is in generating the AI response or in the UI/connection itself.

Slow AI Responses

If the AI takes a long time to respond:

Check your LLM tier:

LLM Tier	Model	Expected Speed
Eco	DeepSeek V3	Fast for simple prompts, slower for complex reasoning
Standard	Gemini	Good balance of speed and quality
Premium	Claude Haiku	Fast with high quality
BYOK	Your chosen model	Depends on the model and provider

Slower responses are normal for complex prompts, long conversations, or models that prioritize accuracy over speed.

What you can try:

• Start a new conversation to reduce context length (long conversations accumulate tokens)
• Use shorter, more focused prompts
• If on Eco tier, consider upgrading to Standard or Premium for faster responses
• If using BYOK, check your provider's status page for outages

High Memory Usage

Each instance tier has a memory limit:

Tier	Memory Limit
Budget	1 GB
Balanced	2 GB
Pro	4 GB

If the container exceeds its memory limit, Docker restarts it automatically. This causes a brief interruption (30-60 seconds) while the gateway restarts.

Signs of memory pressure:

• Instance periodically becomes unresponsive for 30-60 seconds
• Dashboard shows the instance restarting without manual action
• Complex operations (web browsing, large file processing) trigger restarts

What you can try:

• Avoid tasks that load large files into memory
• Limit the number of active skills and plugins
• Upgrade to a higher tier if you regularly hit the memory limit

Container Keeps Restarting

If your instance is in a restart loop:

1. Check the dashboard -- Look for repeated status changes between "Running" and "Deploying"
2. Recent configuration changes -- If you changed openclaw.json or installed a new skill, it may be causing a crash
3. Try rebuilding -- A rebuild creates a fresh container with your configuration. Go to your instance dashboard and use the "Rebuild" option
4. Remove recently added skills -- If the issue started after adding a skill, that skill may be incompatible

Slow Web UI Loading

If the web UI itself (not the AI responses) loads slowly:

• Network distance -- The data center is in Falkenstein, Germany. Users far from Central Europe may experience higher latency.
• Browser extensions -- Some browser extensions interfere with WebSocket connections used by the gateway. Try in an incognito window.
• Port 8080 throttling -- Some ISPs or firewalls throttle traffic on non-standard ports. This is rare but possible.

When to Upgrade Your Tier

Consider upgrading if:

• Your instance hits the memory limit regularly (repeated restarts)
• You run multiple skills that consume significant resources
• You need to process large files or run web automation tasks
• Your workload has grown beyond what the current tier supports

Check the Choose Your Tier page for a comparison of what each tier offers.

Related Docs

• Troubleshooting Common Issues -- General troubleshooting steps
• Instance Monitoring and Health -- How to monitor your instance
• Resource Limits -- Memory, CPU, and storage limits per tier
• Choose Your Tier -- Tier comparison and upgrade options