Use your own API key

You can connect any of these, and the console links below are the providers' own:

• OpenAI: platform.openai.com/api-keys. The API will not answer until the account has a balance.
• Claude (Anthropic): platform.claude.com/settings/keys. Keys start with sk-ant-.
• Gemini (Google): aistudio.google.com/apikey. There is a free tier with rate limits.
• xAI, DeepSeek, Mistral and Groq: each has its own console. Mistral and Groq have free tiers with limits.
• OpenRouter: one key that reaches every model in Copilotly, with one bill. If you have a direct key for a provider, Copilotly uses that one first for that provider's models.

See the model pages for ChatGPT, Claude, DeepSeek and the others to decide which provider you want to start with.

Requests for that provider's models run on your account:

• Every model, no caps. That includes the frontier models, which are otherwise Pro-only and limited to 30 a day, so a Free user can reach them with a key.
• No credits used. You pay the provider's list price and nothing more to them. Copilotly does not mark it up.
• Covers more than chat. Craft, copilots and the reply assistants in Gmail, LinkedIn, X, WhatsApp and similar sites use the key too.

Be clear-eyed about cost. A deep research report or a long file chat on a frontier model can use a lot of tokens, and you are billed for all of it by the provider. Set a spending limit in the provider's console before you start. The key covers chat requests to that provider's models. Image generation and whole-page translation are Copilotly services with their own daily allowances, so do not assume a key lifts those.

This is the question that matters most, so plainly: your key is sent to Copilotly's server and stored there. It is not kept only in your browser.

• It is checked with the provider using a read-only call before it is saved.
• It is encrypted with AES-256-GCM, never sent back to your browser, and only its last four characters are shown.
• It is used only to answer your own requests and is not shared or used for training.
• If a provider rejects the key, Copilotly pauses it and answers for you instead, with a notice.
• When you remove a key it is deleted at once. If you think it leaked, revoke it in the provider's console as well.

If you are not comfortable with a key on a third-party server, create one with a low spending limit just for Copilotly, or do not use this feature. Browser-only extensions that keep the key locally exist, and for some people that is the better trade.

BYOK suits people who already pay for API access, people who want a specific model or provider the free plan limits, and heavy users for whom per-token pricing is cheaper than a subscription. Light users often do better on the free plan or Pro.

It also pairs well with deep research and long file chats if you want them without the daily caps, as long as you keep an eye on provider bills.

A few things to do before you connect a key.

• Set a spending limit in the provider's console, or use prepaid credit, so a runaway prompt cannot cost more than you decided.
• Start with OpenRouter if you want variety. One key reaches every model in the Copilotly picker, with one bill and no markup on model prices. Direct provider keys make sense if you only use one family of models.
• Use direct keys alongside it. A direct key is used first for that provider's models, so you can keep OpenRouter as the fallback for the rest.
• Check the status on each card. A card shows whether the key is working and how many requests it has handled. A paused key means the provider refused it, usually because of billing or a revoked key.
• Know the price of what you ask. Long documents, deep research and frontier models cost far more per message than a short chat on a fast model. Provider dashboards show usage by day.

If you only need a few premium replies a day, the free plan's 10 a day may be enough without any setup. If you want unlimited premium models without managing keys, Pro is the simpler route. Many people use both: Pro for everyday chat and a key for the occasional frontier-model task.

Both work. The right one depends on how many model families you use.

One direct key

If you mostly use one family, a direct key from that provider is the simplest. You see usage in the provider's console, and its own free tier or credit applies. The Gemini API key guide and the Groq quickstart show how short that setup is.

OpenRouter for variety

OpenRouter is a router: one key and one bill that reaches many providers' models. Its quickstart covers creating a key and adding credit. The trade-off is that you depend on a second company between you and the model.

Both together

If you connect a direct key for a provider as well as OpenRouter, Copilotly uses the direct key first for that provider's models, and OpenRouter covers the rest.

Situation	Better choice
You only ever use Claude models	Claude key
You want to try many models, one invoice	OpenRouter
You want a free tier to experiment	Gemini, Mistral or Groq key, within their limits
You need GPT models and nothing else	OpenAI key (balance required)

There is no fixed answer, because the provider charges by tokens and the amount depends on what you ask.

What drives the bill

Long inputs, long outputs and frontier models. A short question to a fast model costs very little. A deep research report, a large file chat or a long conversation on a frontier model costs far more, and the whole conversation is sent again with each message.

How to keep it in check

Set a spending limit or use prepaid credit in the provider's console, check the usage dashboard weekly and use a cheaper model for everyday tasks. Keep the frontier model for the jobs that need it.

When a subscription is cheaper

If you send many long messages a day, a flat plan can cost less than per-token billing. Compare against Pro pricing after a week of real use, or read the ChatGPT comparison and Claude comparison to see where each approach fits.

A key is a password with a price attached. Treat it like one.

Create a key just for Copilotly

Name it clearly in the provider's console. If it is ever abused, you can revoke it without breaking other apps that use a different key.

Put a cap on it

Use prepaid credit, a monthly budget or a hard limit where the provider offers one. With a cap in place, a mistake costs a known amount.

Rotate on a schedule

Create a new key, save it in Copilotly, then delete the old one at the provider. It takes a couple of minutes and limits the damage of a key that has been seen by someone else.

Remove it when you are done

Remove the key from the card in settings, which deletes it from Copilotly at once. Then revoke it at the provider. Doing both means nothing live is left behind.

Never paste a key into a chat

A key belongs in the API Keys page only. If you pasted it into a message by mistake, revoke it and make a new one.

The models available depend on which key you connect, and Copilotly's model picker lists the current set. As of November 2026 it includes OpenAI, Anthropic, Google, xAI, DeepSeek, Mistral, Qwen and Kimi families, plus fast models. A direct key covers its own provider's models. OpenRouter is the way to reach the others with a single key, and Groq hosts a range of open models.

The frontier models

These are the most expensive and are normally a Pro feature, with a 30-a-day cap. With a key, you pay the provider's price and Copilotly does not apply the cap. That is the main reason a free user would connect one. The pages on Claude and ChatGPT cover how those families behave.

Fast models

These are cheap per message, so a key is rarely needed for them. If you mostly use fast models, the free plan may already do what you need.

Chat only

The key covers chat requests. Things that are Copilotly's own services, such as image generation and whole-page translation, keep their own daily allowances. See image generation and page translation for those limits.