If your AI agent has stopped responding to messages, follow these diagnostic steps in order. Most issues are resolved within the first three steps.
Step-by-step diagnosis
Check server status on your Dashboard
Open your OneClickClaw dashboard and look at the server status indicator. A green dot means your server is online and healthy. If you see red or offline, your server needs attention. See the Server Offline article for next steps. A blue Upgrading badge means your agent is moving to a newer version; it is back in about half an hour.
Check that an AI model is connected
Your agent needs your own AI subscription or API key before it can answer. On OpenClaw, open your agent and go to Settings > Models. On Hermes Agent, look at the AI Model tab of your server page. If nothing is connected, see No AI model connected below.
Verify your API key or plan is active
Go to your AI provider's dashboard (e.g., console.anthropic.com) and check that your key is still active and has available credits. If you signed in with a subscription, check that the subscription is still active.
Check the Event Viewer for recent errors
On your Overview page, scroll down to the Event Viewer terminal. Look for recent error events (shown in red). These often reveal the exact cause, such as a failed API call, a configuration issue, or a resource limit.
Try sending a test message on a different channel
If you have multiple channels connected (e.g., Telegram and Discord), send a test message on the other channel. If the agent responds there, the issue is with that one channel connection. See Channel Connection Issues.
Use the Troubleshooting card
On your server page, open the Troubleshooting card. On OpenClaw, press Run next to Run Diagnostics for a full health checklist (gateway, provider key, channels), or next to Test Provider Key to check your key. On Hermes Agent, the card offers Check my dashboard, Check DNS, Restart the dashboard and Restart messaging.
Reboot the server
At the top of your server page, click Reboot. This restarts your whole server, including your agent, and clears stuck states. The server is briefly unavailable while it restarts.
Open a support ticket
If none of the above resolves the issue, click Stuck? Get help at the top of your server page and tell us what you see, when it started, and what the Troubleshooting card reported.
Step 1: Dashboard
Steps 2-3: AI model
Step 4: Event Viewer
Step 5: Test Channel
Steps 6-7: Troubleshooting card and Reboot
Step 8: Support
Common causes
- No AI model connected: The most common cause by far. The server is healthy, but the agent has no AI to think with. See below.
- Expired or revoked API key: Your AI provider key was regenerated or ran out of credits. Connect a new key inside your agent (OpenClaw: Settings > Models > Connect provider; Hermes Agent: the AI Model tab or your Hermes dashboard).
- Provider rate limit: You hit the requests-per-minute limit on your AI provider. Wait a few minutes or upgrade your provider plan.
- Server memory full: The server ran out of RAM and the agent process was stopped. Reboot the server and consider upgrading your plan.
- Disk full: A full disk can stop your agent from saving anything. See High Resource Usage.
No AI model connected
Every plan is bring-your-own-AI, so a new server has no AI model until you connect one. Until then the agent stays silent, even though the dashboard shows the server as healthy.
- OpenClaw: open your agent, go to Settings > Models and press Connect provider. For a ChatGPT plan choose OpenAI > Codex login (device code) (older servers may still call it ChatGPT Device Pairing). This needs ChatGPT Plus or higher, and you must first turn on Enable device code sign-in in your ChatGPT settings (Settings > Security and login > App security). For Grok choose xAI (Grok) > xAI OAuth, which needs SuperGrok or X Premium+. Or paste an API key. Then pick the provider in the Default model list.
- Hermes Agent: open the AI Model tab of your server page and use Use a subscription you already have (a ChatGPT plan, SuperGrok or X Premium+, Nous Portal or a MiniMax plan) or an API key.
Step-by-step guide: Connect your ChatGPT plan.
Tip
If your agent answers with an error that your plan does not support a model, choose another model: on OpenClaw in the Default model list, on Hermes Agent with Change Model on the AI Model tab. A ChatGPT plan does not serve every model on every tier.
Still not working?
If you have gone through all the steps and your agent is still not responding, open a support ticket from your dashboard. Include what the Troubleshooting card reported and a description of when the problem started. Our support team responds within the response target of your plan.
Note
Support response times: Starter (12 hours), Hobbyist (12 hours), Pro (4 hours).
Related articles: API Key Security, Server Status Indicators