Fix AI connection problems
Restore a connection when a key, sign-in, local server, or model is not working.
Written By Sujeevan Nagarajah
Last updated 22 days ago
Start with the error shown in Vinaa: it often includes the provider's reason. For a cloud connection, also check that your Mac has internet access.
Find the connection that needs attention
Choose Vinaa → Settings → AI Models.
Open the affected connection.
Read its status or error.
If an issue indicator appears at the bottom of the Folder Panel, click it to open Issues. Fix opens the affected connection's settings.
Setup required means a ChatGPT Plan or Claude Plan connection needs setup on this Mac. A connection synced from another Mac does not bring all of that local setup with it.
Replace an invalid API key
An API key is a secret credential for your provider account.
Open the connection's Edit… sheet.
Replace API Key with a complete, valid key from the matching provider account.
Click Save.
Check your provider's credits, billing, model permissions, and usage limits if the key works but messages fail. For OpenRouter browser sign-in, use Sign In Again in Edit….
Restore ChatGPT Plan sign-in
If the status reports that you are not logged in or your session expired, open Terminal and run codex login. Complete OpenAI's browser sign-in, then return to Vinaa.
For Folder access required, click Grant Access to ~/.codex on the connection page. A successful setup shows Authentication verified.
Restore Claude Plan sign-in
Claude Plan is Experimental. Follow the connection's status:
After fixing the problem, click Check Again. Accept the displayed Anthropic terms on each Mac. See the official sign-in guide and installation and update guide.
Check a local server
Ollama: keep Ollama running and download a model first. This Mac normally uses
http://localhost:11434. Ollama Cloud requires a key.LM Studio: start its server from Developer and load a model. The default Base URL in Vinaa is
http://localhost:1234. Supply API Token if authentication is enabled.
Correct the address or credential through Edit…, then Save.
Check feature access and refresh models
A free API tier may allow chat while restricting tools or particular models. For example, Google Search availability varies by Gemini model and tier; check Google's pricing table. Try disabling Web Search in the tools panel (⇧⌘T), or choose a supported model with ⌘L.
After fixing a connection, click its Refresh button. Sync on the main AI Models page refreshes across connections. Use Choose Models… to select additional models.
Remove a connection you no longer need
Click Delete Connection… at the bottom of its page. If settings still use its models, choose Migrate all to… to select a replacement, or Clear all, then Delete Connection. When no settings depend on it, confirm Delete instead.
Related articles
Connect an AI provider
Use your ChatGPT plan in Vinaa
Use your Claude plan in Vinaa
Report a problem
Was this helpful?
More in Connect Your AI
Which AI connection should I use?Get an OpenAI API keyGet an Anthropic (Claude) API keyGet a Google Gemini API keyStill need help? Ask the team