MorseSign inStart freeOpen Morse

What is BYOK AI? Bring your own key, and whose rules then cover your prompts

The Morse teamPublished 9 min read

BYOK AI, short for "bring your own key," is an AI app that runs on an API key you get from OpenAI, Anthropic, Google, or another provider, so the provider bills you directly and your prompts fall under that provider's API terms rather than a consumer plan's. As of November 2026, it doesn't take the app out of the middle. Some BYOK apps send your prompts straight from your browser to the provider, while others, including Cursor and Warp, pass every prompt and your key through their own servers first.

We read the bring-your-own-key pages of TypingMind, T3 Chat, OpenRouter, Cursor, Warp, and Visual Studio Code on November 20, 2026, along with the API pages of OpenAI and Google, to see where the key goes and whose rules cover what you type. Warp and Cursor deserve credit for saying plainly that their zero data retention deals stop applying the moment you switch to your own key.

Morse is a private AI chat app with GPT, Claude, Gemini, DeepSeek, and uncensored models, encrypted so even Morse can't read your chats. Like a BYOK app, it reaches 80+ models, open and closed, in one account. Unlike one, Morse sends your message under its own account, so the model provider never learns your name, email, account, or IP address, and no AI provider trains on what you send through Morse.

What does BYOK mean in AI?

It means you supply the API key and the app supplies the interface. An app with BYOK lets you paste a key from a model provider, then sends your chats to that provider on your account instead of on the app's own. Cursor's help page says "Your provider (OpenAI, Anthropic, Google, and others) bills you directly for the model cost of those requests." TypingMind is built entirely on it, and its FAQ says "You will need a working AI API key" because "TypingMind does not sell API credits."

The letters also have a second meaning. In enterprise security, BYOK means encrypting the data a vendor stores with a key the customer manages. OpenAI's Enterprise Key Management page describes this for ChatGPT Enterprise and the API, and says OpenAI "supports Bring Your Own Key (BYOK) encryption" with AWS, Google Cloud, Azure, and Oracle key vaults. That kind of BYOK doesn't hide your content from the vendor. The same page says that on decryption, OpenAI asks your cloud key service to unwrap the data key, "and we decrypt the data with the DEK." The customer decides who can unlock stored data and when, and the vendor still reads it while working.

Where does a BYOK app keep your API key?

Wherever the app chooses, and the five apps in the table give five different answers. Each row comes from the app's own pages, read on November 20, 2026.

AppWhere your key is keptDoes your prompt pass through the app's servers?Does every request use your key?
TypingMindIn your browser, unless you turn on Cloud Sync, which stores "API Keys and License key" on TypingMind's server (Cloud Sync page); optional AES encryption behind a password you set (FAQ)No by default: requests go "directly from your device's browser to the AI provider's server," with an opt-in proxy for providers that block browser requestsYes for chat, since TypingMind sells no API credits
Warp"Your device's secure storage"Yes: "your prompt, context, and key pass through Warp's servers," and Warp "does not retain your personal API key"No: Auto models use Warp's own inference "even if you've configured your own API keys," and so does Codebase Context
Cursor"Not stored on our servers," but "sent to our backend with every request"Yes: "all requests are routed through Cursor's servers for final prompt building"No: "Tab completion continues using Cursor's built-in models"
OpenRouterOn OpenRouter, "securely encrypted"Yes, since OpenRouter routes the request to the providerNot always: prioritized keys are tried first, then OpenRouter's shared endpoints unless you block them
VS CodeNot stated for every provider; the docs say to use an input variable "instead of committing a raw key"Not stated on the page we readNo: titles, commit messages, and intent detection use "built-in utility models provided by GitHub Copilot" by default

A key kept "on your device" can still travel with every request, as Cursor's does, so ask whether the key leaves your machine as well as where it's saved. The model you pick in a BYOK app also isn't always the model doing all the work. Cursor's privacy settings in more depth are in is Cursor private, and OpenRouter's fallback rules in Morse vs OpenRouter.

Can a BYOK app see your prompts?

Yes, whenever the app's servers sit between you and the provider. TypingMind's privacy policy, last updated September 19, 2026, says it has no "intermediate server to look at the requests or log the responses" in its default mode, and that its optional proxy will "simply pass the request through as-is." That design keeps TypingMind out of the path, and it's the clearest statement of the kind we found.

Warp and Cursor work the other way, because they build the prompt on their servers. Warp's BYOK page says your prompts and responses "pass through Warp's servers" and that Warp "applies your account's privacy settings to this content," including its setting to store AI conversations in the cloud. Cursor's data use page says that "Even if you use your API key, your requests will still go through our backend." So BYOK changes who bills you. Whether the app sees your words depends on how it's built, and the app's own history settings still apply to anything it keeps.

Whose privacy terms cover your prompts with BYOK?

Your provider's, for your account, and several apps say so in their own pages. TypingMind's policy says "Data sent to any AI providers is governed by their respective privacy policies." Cursor's API key page says its "Zero Data Retention policy does not apply when you use your own API keys," and Warp's says "Warp cannot enforce provider ZDR for BYOK." T3 Chat's FAQ puts it in bold: if you bring your own key, "you are responsible for ensuring that your organization has configured their settings to not allow training."

OpenAI's platform data controls page, for one, says API data hasn't trained its models since March 1, 2023 unless you opt in, but abuse-monitoring logs that "may contain certain customer content, such as prompts and responses" are kept "for up to 30 days." A free key can come with different rules. Google's Gemini API terms, effective March 23, 2026, say that on unpaid services "human reviewers may read, annotate, and process your API input and output," and "Do not submit sensitive, confidential, or personal information to the Unpaid Services." A free Gemini key in a BYOK app puts your chats under those rules.

OpenRouter adds a wrinkle of its own. You can mark a BYOK key as covered by zero data retention so routing treats it that way, but its guide says "The declaration is your statement about your own provider agreement. OpenRouter does not check it." We compare six companies' API terms side by side in is the OpenAI API private, and explain the term itself in what is zero data retention.

Does a BYOK request still identify you?

Yes, to the provider. OpenAI's key safety page says "An API key is a unique code that identifies your requests to the API," and the key belongs to an account you opened with the provider and pay for. Whatever the app does with your history, every prompt reaches the provider tied to that account, including the prompts it keeps in abuse-monitoring logs. We cover those logs in what is AI abuse monitoring.

Is BYOK safe, and is it allowed?

It can be, if you treat the key like a card number. OpenAI's key safety page says "Your API key is intended to be used by you. The sharing of API keys is against the Terms of Use." The same page warns, "Never deploy your key in client-side environments like browsers or mobile apps," because others could "make requests on your behalf." That advice is written for developers shipping apps to strangers. A BYOK chat app in your own browser is a different case, but the risk it names is real if the key leaks.

What OpenAI's page recommends works for any key you paste into an app:

  1. Set an expiry date. "Project keys can be created with an expiration date after which any requests from those keys will be rejected."
  2. Set a hard spend limit. OpenAI suggests monthly spend limits and says to "enable hard enforcement," and notes that "Spend alerts alone don't stop API traffic."
  3. Use one key per app, so you can revoke one without breaking the rest.
  4. Encrypt the key where the app allows it. In TypingMind, that's API key settings, then Encrypt API key, with a password.
  5. Check your provider's training setting, since the app won't do it for you.

Is BYOK cheaper than a subscription?

Sometimes, and the fees are easy to miss. OpenRouter's BYOK guide says using your own keys costs "5% of what the same model/provider would cost normally on OpenRouter," after a free allowance that on pay-as-you-go "includes $25,000 per month." Cursor's page on bringing your own API key says own-key requests on individual plans "don't draw from your included usage," but its Teams and Enterprise plans add a Cursor Token Rate "of $0.25 per million tokens" on eligible models, "including requests made with your own key." TypingMind charges for a license on top of every API bill, which we price in TypingMind alternatives.

How Morse reaches many models without your name on the request

With your own key, every prompt reaches the provider under an API account you opened and pay for, and in many apps it also passes through the app's servers on the way. Turning off training, setting retention, and checking zero data retention all fall to you. Morse is built so the account the provider sees is Morse's, and you choose who can read each chat with a privacy mode per chat:

  • Anonymous, for GPT, Claude, and Gemini without an API account in your name. Morse sends your message under its own account, so the model provider never learns your name, email, account, or IP address. The provider still reads the message to answer it, without knowing it's you, and may keep it for a limited time to check for abuse, under its own policy. So leave out names and details that point back to you.
  • Private, for open-weight models and Grok without a retention setting to chase. New chats start here. Morse has zero data retention agreements with its inference, search, and other providers, so they've agreed to keep nothing, and where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't.
  • E2EE, for prompts you want sealed before they leave your browser. Your browser checks the enclave's hardware attestation, then encrypts the message so only that verified enclave can read it. Morse's servers relay text they can't read. Every E2EE reply has a lock you can open to see each check that passed, and E2EE chats come with Pro. For web search, use a Private chat.

Your chat history is stored in your browser, encrypted, not on Morse's servers, and Morse's servers keep no log of your messages. Morse doesn't train AI on your chats, sell your data, or show you ads. Because Morse doesn't store chats, it usually has very little to give authorities. Save a backup to a file from Settings > Data, since clearing your browser or losing the device deletes chats Morse can't recover.

Morse charges one plan price instead of a bill from each provider. Morse's plans start with a free plan with no card, and Pro is $19 a month ($180 a year) with unlimited messages on Pro models and your choice of model. Developers can use the private inference API.

Start free with no card, or send a first message from the Morse home page without an account.

Sources

Read November 20, 2026.

Common questions

What does BYOK stand for in AI?

Bring your own key. You paste an API key from OpenAI, Anthropic, Google, or another provider into an app, and the app runs the models on your account, so the provider bills you directly. In enterprise security the same letters mean something else: encrypting data a vendor stores with a key you manage.

Does BYOK mean the app can't see my prompts?

Only in some apps. TypingMind sends requests straight from your browser to the provider by default, while Warp and Cursor route every prompt, and your key, through their own servers on the way.

Is BYOK more private than ChatGPT?

On training, often. OpenAI says data sent to its API hasn't been used to train its models since March 1, 2023, unless you opt in. The API still keeps prompts in abuse-monitoring logs for up to 30 days, and every request is tied to your API account.

Does using my own API key turn off training?

It depends on your provider account, not the app. T3 Chat's FAQ says BYOK users are responsible for their own organization's training settings, and on Google's free Gemini API tier, human reviewers may read what you send.

Is it against OpenAI's terms to put my key in an app?

OpenAI's key safety page says your key 'is intended to be used by you' and that sharing keys is against its terms of use. It also warns against putting a key in browsers or mobile apps, so set an expiry date and a hard spend limit on any key you paste into one.

Is BYOK cheaper than a subscription?

Sometimes, but not always free of fees. OpenRouter charges 5% of the normal cost once you pass $25,000 a month of BYOK use on pay-as-you-go, and Cursor's Teams and Enterprise plans add $0.25 per million tokens on many models even with your own key.

Does the model provider know it's me when I use Morse?

Not in Anonymous mode. Morse sends your message under its own account, so the provider never learns your name, email, account, or IP address. It still reads the message to answer, and may keep it for a limited time to check for abuse, under its own policy.

Ask what you actually want to ask.

80+ models, characters, and images, with every chat encrypted and kept on your device. Free to start.

Start a private chat

Read next