Skip to content

Answers

Frequently asked questions

Whether it is free, how keys are handled, what the API costs, which model to pick, and what to do when something breaks.

15 questions · 4 topics

Answers reflect the site as of October 2026.

Free & funding

The honest answers about price, ads and model billing.

Chatting on KeyChat has no site fee. No signup or subscription is required and there are no per-message credits. Official character licences and extra server storage are optional purchases.

The models themselves are not free. Requests go to the provider with your own API key, and the provider bills that usage to your account. KeyChat never touches that money.

In short: the site is free, the model runs on your key.

Running the site costs money — a domain, hosting, and the time to build and maintain it. Static hosting is cheap, but not zero.

Google AdSense covers those costs. That is why there is no usage fee and no paid tier that locks features away.

Ads are labelled and kept separate from content. You never need to click one, and chat keeps working with an ad blocker enabled.

You do. Usage is billed by each provider to the payment method on your own account.

KeyChat never sees payment details and adds no markup. Whatever your provider invoice says is the whole cost.

The cost figure shown in a conversation is an estimate based on public list prices. Your actual bill depends on provider pricing, caching discounts and currency.

Security & privacy

Where your key goes and where your data stays.

Your key is stored only in this browser. KeyChat has no server that accepts keys, and there is no path for a key to reach the site operator.

Requests go from your browser straight to the provider API. No KeyChat server sits in between.

The limits are real, though. A key can be left behind on a shared device, and browser extensions can read page storage. On a shared machine, switch the storage mode to 'Until this tab closes' or 'Forget on reload' and delete the saved key when you are done.

In IndexedDB, inside this browser. Even when signed in, chats are never uploaded or synced to the account.

That means chats do not appear on another device or in another browser, and clearing browser data deletes them.

Export important conversations as JSON from Settings. The file contains the conversation text, so keep it somewhere safe. API keys are never included.

Revoke the old key in the provider console, create a new one, and paste the new key into KeyChat settings.

KeyChat does not keep a copy of your key, so it cannot recover or revoke it for you. If you suspect a leak, revoke it in the provider console right away.

KeyChat sets no cookies of its own. Settings and chats live in browser storage (localStorage, IndexedDB), not in cookies.

Third-party ads such as Google AdSense may use cookies to serve ads. The privacy policy explains how to turn off personalised ads.

Using KeyChat

Model choice, mobile, backups and local models.

To start on a free tier, Gemini 3.8 Flash is the easy choice — Google AI Studio gives you a free quota with no card required.

For value, Claude Haiku 5.5, GPT-6 Luna and DeepSeek Flash work well. For roleplay, long writing and code, Claude Sonnet 5.5 or Opus 5.5 are steady.

If you want to move between models, one OpenRouter key covers most of them. There is no single right answer — try the same question on two models.

Yes. Open the same address in a mobile browser and you get the same app.

Mobile browsers sometimes reclaim storage, so you may need to re-enter your key. Adding the site to your home screen makes it open like an app.

Yes. Demo mode lets you explore the interface before committing. Real model answers require a key.

If you do not have one yet, the setup guide walks through issuing a key.

Settings lets you export every conversation to a JSON file and import it again on another device.

There is no server sync — moving the file yourself is the only way to transfer history.

Yes. Choose the custom provider and enter an OpenAI-compatible endpoint — for example http://localhost:11434/v1 with the model llama3.2.

Ollama blocks browser requests by default, so you may need to set OLLAMA_ORIGINS. The setup guide has the exact steps.

Troubleshooting

What to check when a request fails.

Calls go from your browser straight to the provider, so a provider or network policy can block the request.

The reliable fix is an OpenRouter key — OpenRouter officially supports direct browser calls.

Browser extensions and corporate proxies can also be the cause. Try again in a private window.

Check for leading or trailing spaces and line breaks from the copy.

Make sure the provider matches the key. Anthropic keys (sk-ant-) only work with Claude, OpenAI keys (sk-) with OpenAI, Google keys (AIza) with Gemini.

A freshly created key can take a moment to become active on the provider side.

You have hit a free quota or a per-minute request limit. Wait a moment and retry.

If it persists, check usage and billing status in the provider console.

Did not find an answer?

Send an email. Bug reports, feature requests and ad enquiries all go to the same address.