Troubleshooting

This page covers the most common issues users encounter on The Chat Agent. Find your situation below; if the problem persists, contact support.

Provider API key errors

Invalid or expired API key

If a chat returns an error such as "Authentication failed" or "Invalid API key", the key you have stored for that provider is incorrect, expired, or has been revoked. Each provider uses its own key format — an OpenAI key will not work for Anthropic or Google Gemini.

To fix this, open Settings → Providers, select the affected provider, and update the API key. You can generate or rotate keys directly on the provider's platform (for example, the OpenAI dashboard or Anthropic console). See Providers for step-by-step instructions per provider.

Insufficient credits on your provider account

Some providers return an error such as "Insufficient quota" or "You have exceeded your current quota" when your provider account has run out of credits or hit a spending limit. This is separate from your Chat Agent subscription — you need to top up your balance or raise your spending limit directly on the provider's platform. See Providers for links to each provider's billing settings.

A model is missing from the dropdown

Provider not connected

Models only appear in the model selector once their provider is connected. If a provider is missing from Settings → Providers, add it and enter a valid API key. Once saved, all available models for that provider become selectable across new and existing chats.

Model not available on your provider plan

Some provider accounts restrict access to certain models — for example, access to GPT-4o requires prior usage history on OpenAI, and Claude models may require an approved Anthropic account. Confirm the model is enabled for your API key on the provider's platform, then reload the page to refresh the model list. See Providers and Models for per-provider details.

File upload fails

Unsupported file type or file too large

The Chat Agent accepts a specific set of file types and enforces a maximum file size per upload. If your upload is rejected, verify that the file extension is supported and that the file does not exceed the size limit. See File Uploads for the full list of accepted types and current size limits.

Model does not support attachments

Not all models accept file attachments. If you attach a file and the model ignores it or returns an error, switch to a model that supports vision or document inputs — such as GPT-4o or Claude Sonnet. The model picker displays an icon next to models that accept attachments. See File Uploads for a list of attachment-capable models.

Share link issues

Shared link returns an error or "not found"

Share links can be revoked by the conversation owner at any time. If a link you received no longer works, ask the owner to re-share the conversation and send you the updated link. See Sharing for instructions on managing share links.

Recipient sees a read-only view

Shared conversations are intentionally read-only for recipients. A recipient can read the full conversation thread but cannot send new messages or modify it. Only the conversation owner or team members with edit access can participate. This is by design — see Sharing for details on share link permissions.

Message limits and billing

Reached the free-tier message cap

The free plan includes a monthly message allowance. When you reach the cap, new messages are paused until the next billing cycle resets your usage, or until you upgrade. Paid plans raise the monthly message cap and unlock additional features including team workspaces, agent builders, and workflow automation. Visit Pricing to compare plans and upgrade.

Login and account issues

Forgot password or can't log in

Use the Forgot password link on the login page to request a password reset email. Check your spam or junk folder if the email does not arrive within a few minutes. For further account help, see Account.

Verification email not received

After registering, a verification email is sent to the address you provided. If it does not arrive, check spam, then use the Resend verification option on the login page. If the problem persists, see Account settings or contact support.

Slow or interrupted responses

Slow streaming or mid-response cutoffs are usually caused by one of the following:

• Network conditions — a weak or unstable connection can interrupt the response stream. Refresh the page and send the message again.
• Provider latency — AI providers occasionally experience elevated latency or partial outages. Check the provider's status page and try again in a few minutes.
• Model load — large models or long context windows take longer to begin streaming. Switching to a faster model (such as GPT-4o mini or Claude Haiku) for exploratory tasks can significantly improve response speed.