What to do when Ask AI does not answer

"Ask AI runs on your own AI provider for now"

You have no text provider connected that Ask AI can run on. Connect one in Settings, under AI providers: OpenRouter in one click, or a key from one of the providers in Which AI providers and models Ask AI can use. The model menu has a Connect an AI provider link that goes straight there. An image, video or voice provider does not count, and neither does Omne AI for now.

"… has no chat model Omne can use. Pick another provider."

The provider is connected, but its model list had nothing that chats once image, voice and embedding models were left out. Open the model menu and pick another provider, or remove this connection and connect it again with a key that has access to chat models.

"Ask AI can't run on this provider yet"

You chose a provider that can be connected but that Ask AI does not support for chat, such as Amazon Bedrock. Pick another in the model menu.

The provider refused the request

The chat shows the provider's own message, for example about an invalid key, a missing balance or a rate limit. Open AI providers, check the connection, and compare with your usage on the provider's site. A key that worked when you connected can later be revoked, run out of credit or hit a limit. Making a new key and connecting it again fixes the first two.

"That's a lot of messages in a minute"

Omne allows 30 messages a minute per person. Wait a short while and send the message again.

"The answer stopped. Try again."

The connection dropped before the answer finished, or the provider ended the stream. Send the message again. If it happens every time on one model, switch models. A very long conversation can also be heavy: start a New chat and paste in only what you need.

"Stopped after many steps. Ask again to carry on."

A chat that keeps calling tools stops after 16 rounds so it cannot loop forever. Ask a narrower question, or send a message such as "continue" and it will pick up from where it was.

A tool shows a cross

A cross beside a tool line means that call failed. Hover over it to see why. Common reasons:

• It needs reconnecting in Integrations. The account's sign-in expired or was revoked. Open Integrations and use Reconnect on that account.
• It no longer accepts Omne's sign-in. The service rejected the token. Reconnect the account.
• It took too long to answer or it didn't answer. The service was slow or down. Try again in a moment.

If an account could not be reached at all, the AI says so in its answer.

The AI does not use an account you connected

Check three things. First, the account must bring tools, which means it is connected through the maker's MCP server or is GitHub. Second, it must be ticked in the Connectors menu. Third, the provider must take tools: Perplexity does not. Naming the account in your question also helps. See Use your connected accounts in a chat.

Nothing happens after you press Allow

An approval is answered once. If the card already says Allowed or Declined, it is done. If you waited longer than ten minutes the request counted as declined. Ask the AI to try again and it will show a new card. See How approvals work.

The menu does not show a provider you just connected

The model menu reads your providers again each time you open it. If a provider is still missing, check that it is a text provider and that it appears under Connected in AI providers.