FAQ & troubleshooting
Answers to common questions about Cordy for Chrome — cost, browsers, the consent dialog, re-granting access, WebGPU and local models, provider detection, and the full keyboard-shortcut list.
Common questions and fixes. If something isn't here, the feature pages linked at the bottom go deeper.
General
Is Cordy free?
The extension is free. Because it's bring-your-own-key, you pay your AI provider directly for what you use — or nothing at all if you run local, on-device models. There's no Cordy subscription or account.
Is Cordy on the Chrome Web Store?
Not yet. While the store listing is prepared, you install Cordy as an unpacked/developer build or from a self-distributed package. See Installation.
Which browsers and versions are supported?
Chrome 142 or newer, and Chromium browsers such as Microsoft Edge at an equivalent version. Cordy refuses to install on older builds rather than break at runtime.
Wasn't this called "Web Nexus"?
Yes — Cordy for Chrome was formerly Web Nexus. It was renamed to Cordy across the whole product. Older references to "Web Nexus" mean this extension.
Is my data safe?
Yes. Everything is stored locally, keys are encrypted, prompts go straight to the provider you configured, and there's no telemetry and no Cordy backend. The full model is in Privacy & permissions.
Setup and providers
Why did I see a "Send to a cloud AI service?" dialog?
That's the one-time cloud consent gate. The first time you send anything to a cloud model, Cordy confirms that your text goes to the provider you configured (and nowhere else). Accept it once; you can withdraw consent later in Settings → Data & privacy. Local models never show it. See Privacy.
It says "AI feature not configured" or lists no models.
No provider is connected yet. Open the dashboard → Settings → AI Config → API Providers, add your key, and click Detect. Then pick a Chat Model under Scenario Model Routing. See Providers & models.
A provider shows "Unavailable" or isn't detected.
Re-check the API key and Base URL, then click Detect again. Confirm the key is active with the provider and that you have network access. For a custom OpenAI-compatible endpoint, remember to type its model IDs manually — custom endpoints aren't auto-discovered.
Do I need an API key at all?
No. You can run entirely on-device with Chrome's Gemini Nano, in-browser Gemma, or a local runtime. See Local & on-device AI.
Site access and permissions
In-page features stopped working after an update.
A major upgrade can reset the optional site-access grant. Re-enable it once in Settings → General → In-Page Features. Chat and panel features are unaffected; only in-page features (selection menu, page read-aloud, the command palette, reading page content) need it.
Why didn't Cordy warn me it can "read all my data on all sites"?
Because it doesn't request that at install. Site access is optional and granted from a button when you choose to use in-page features. See Privacy & permissions.
How do I revoke access?
- Site access: Settings → General → In-Page Features → Revoke.
- History / bookmarks: Chrome's extension settings.
- Cloud consent: Settings → Data & privacy → Withdraw consent.
Local models and WebGPU
What do local models need?
Turn on Local AI (Experimental) in Settings and accept the risk notice. In-browser models (Gemma) and Kokoro TTS require WebGPU; Gemini Nano requires enabling Chrome's AI flags. See Local & on-device AI.
"WebGPU is required but not available."
Your browser/device doesn't expose WebGPU. Update Chrome and your graphics drivers. If it still isn't available, use Chrome built-in AI or a local-native runtime instead of the in-browser models.
A model download is stuck or failed.
Cancel and retry, and make sure you have enough free disk space (the Gemma models are several GB). A stalled download can also mean WebGPU isn't available.
How do I enable Chrome's Gemini Nano?
Set chrome://flags/#prompt-api-for-gemini-nano to Enabled and chrome://flags/#optimization-guide-on-device-model to Enabled BypassPerfRequirement, relaunch Chrome, then use Initialize Model in Settings → Local AI.
Chat
How do I clear or export a conversation?
Use /clear to archive the current chat and start fresh, or /export to save it as Markdown. You can also copy any single message. See Chat.
My attachment was rejected.
Attachments are limited to 5 files, 10 MB each, and need a multimodal cloud model. Local models don't accept attachments — switch to a cloud model or remove the files.
The reply timed out.
The connection to the background worker was interrupted. Send again; if it keeps happening, reload the extension from the panel's error screen.
Keyboard shortcuts
Global (managed by Chrome, remappable at chrome://extensions/shortcuts)
| Action | Windows / Linux | macOS |
|---|---|---|
| Toggle the side panel | Ctrl + Shift + Y | Cmd + Shift + Y |
| Open the command palette | Ctrl + Shift + P | Cmd + Shift + P |
| Collect the current window's tabs | Ctrl + Shift + K | Cmd + Shift + K |
| Open the extension action | Ctrl + Shift + L | Ctrl + Shift + L |
Chat and conversation tabs
| Action | Shortcut |
|---|---|
| Send message | Ctrl + Enter (default) or Enter (configurable) |
| New line | Enter or Shift + Enter (matches your send-key choice) |
| Browse input history | ↑ / ↓ in an empty composer |
| New conversation tab | Alt + T |
| Close current tab | Alt + W |
| Reopen last closed tab | Alt + Shift + T |
| Jump to tab 1–9 | Alt + 1 … Alt + 9 |
| Next / previous tab | Ctrl/Cmd + Tab / Ctrl/Cmd + Shift + Tab |
| Open command menu (in composer) | / |
| Insert a browser-context mention | @ |
You can change the send-message key in Settings → Shortcuts.