Troubleshooting and FAQ · AI connection problems
Documentation

AI connection problems

What to do when the AI is slow, rate-limited, can't sign in, or refuses a request.

Most AI hiccups pass on their own. EaseKit retries automatically when a provider is rate-limited, overloaded or the connection drops. When something needs your attention, the chat shows what happened with a button to fix it: Try Again, Set Up in Settings, Sign in to Claude, Sign in with ChatGPT or Subscription Settings.

Troubleshooting

If this happensTry this
“No provider” / Set Up in SettingsConnect one in Settings → Providers. See Connect an AI provider.
Subscription limit reachedWait for your plan to reset, or turn on Continue with the other plan at the limit if you have both Claude Code and Codex.
CLI not foundClick Install in Terminal in Settings → Providers. If typing into Terminal isn't allowed, EaseKit copies the command instead; paste it with ⌘V and press Return.
Sign-in didn't finishClick Sign in again and confirm in the browser window that opens.
The key is rejectedPaste it again and click Save. Check that it's an Anthropic key for Claude and an OpenAI key for OpenAI.
Responses get cut offRaise Max response length in Settings → Providers → Responses.
A long session feels forgetfulRaise History in context, or start a New Session in Same Style (⇧⌘N).
Answers are slowSwitch AI power to Eco, or pick a faster model with ⌥⌘M.
A request is declinedRephrase it. For Claude models that support it, Claude: refusal fallback (server-side) lets another model take over automatically.