# Morse: the full text, for AI assistants > Morse (trymorse.ai): Private AI chat with GPT, Claude, Gemini, DeepSeek, and uncensored models, encrypted so even Morse can't read your chats. This file collects Morse's key pages as plain markdown: the app, its privacy tiers, how encryption and verification work, the plans, and the developer API. Last updated November 17, 2026. The short version is https://trymorse.ai/llms.txt; the maintained briefing for assistants is https://trymorse.ai/ai-info. ## What Morse is Morse (also called Morse AI) is a web app for chatting with 80+ AI models in one account: GPT, Claude, Gemini, Grok, DeepSeek, Kimi, Qwen, GLM, and more, plus image models. It runs in the browser on any device and installs from the browser; there's no native iPhone or Android app yet. Three things set it apart: - **Private by default.** Morse can't read your messages. Your chat history is encrypted in your browser and kept on your device, not on Morse's servers. - **Uncensored models on paid plans.** Pro and Max include uncensored open models that answer instead of refusing, such as plain medical, legal, and security answers. Nothing on Morse helps with anything illegal. - **Every top model in one app.** One subscription covers frontier models and open-weight models, with no ads on any plan. ## The three privacy tiers Every tier hides who you are from the model provider. You pick the tier for each chat. | | Anonymous | Private | E2EE | | --- | --- | --- | --- | | Your message, hidden from Morse | Yes | Yes | Yes | | Who you are, hidden from the model provider | Yes | Yes | Yes | | Your message, hidden from the model provider | No | No | Yes | | Nothing kept by the model provider | No | Yes | Yes | | Guaranteed by hardware, not a contract | No | No | Yes | | Web search in the app | Yes | Yes | No | | Models like GPT, Claude, and Gemini | Yes | No | No | - **Anonymous.** The model provider reads your message but never learns who you are, and may keep it briefly to check for abuse. This is the tier for GPT, Claude, and Gemini. - **Private.** The model provider reads your message but never learns who you are, and has signed a zero data retention agreement to store none of your messages. Open-weight models like DeepSeek, Kimi, and GLM, plus Grok. - **E2EE.** Your browser encrypts your message so only the model provider's secure enclave can decrypt it. Open-weight models only; no web search or image generation in E2EE. ## How end-to-end encryption works 1. **Your browser checks both enclaves.** Each enclave's key comes in a report signed by the hardware, proving it belongs to a genuine enclave running the expected software. 2. **Your browser encrypts your message twice.** The outer layer, for Morse's enclave, holds only what Morse needs to route and bill the message, like the model you picked. The inner layer holds your message, for the model provider's enclave. 3. **Morse forwards your message** to the model provider's enclave, still encrypted, without being able to read it. 4. **The model replies** inside the enclave, which encrypts the reply for your browser alone. 5. **Your browser decrypts the reply** and saves the chat on your device, encrypted. ## How your chats are stored Only your device can decrypt your chats, and only while you're signed in: decrypting takes your browser's key and a key Morse keeps, together. Neither works alone. Each message and file is encrypted separately. What Morse stores: your account (email and sign-in details), billing (plan, balance, usage), a hash of your IP address to stop spam, and one of your two keys, which is useless without your browser's key. What Morse never sees: your messages, your chats (kept only on your devices, encrypted), chats you sync to another device (encrypted end to end on the way), and logs of your chats (Morse's servers keep none). ## Verify it yourself One command checks that Morse's enclaves run the open-source code Morse published on GitHub: ``` npx @askmorse/verify inference.trymorse.ai ``` The check ties each enclave to a commit in four steps: AWS's hardware signs each enclave's measurement; Morse's CI signs a list of the images its enclaves may run, recorded in a public log; a commit to https://github.com/trymorse/morse-enclave names each image's measurement; and each release builds reproducibly, so you can rebuild it and compare. You can run the check in your own CI so a deploy stops if Morse's enclaves ever run an image that isn't on the signed list. ## Plans (checked November 17, 2026) No ads on any plan. Cancel anytime. | Plan | Price | What's in it | | --- | --- | --- | | Free | $0 | 15 messages a day, 3 images a day, Anonymous and Private chats, encrypted on your device, artifacts, web search, and dictation. Needs a real email address. | | Pro | $19 a month, or $17 a month billed yearly | Unlimited messages on Pro models, uncensored text and image models, end-to-end encrypted chats, deep research, sync across 3 devices, plugins and scheduled tasks, 250 credits a month for frontier models | | Max | $49 a month, or $44 a month billed yearly | Everything in Pro, 5,500 credits a month, sync across 5 devices, priority support | 100 credits cost $1. ## Morse Private Inference (the developer API) Morse Private Inference is an OpenAI-compatible API for 80+ open-weight chat models, from DeepSeek, Kimi, and GLM to Qwen, MiniMax, and gpt-oss. - **Base URL:** `https://inference.trymorse.ai/v1`. It speaks OpenAI's Chat Completions format: change the base URL, the key, and the model id, and your OpenAI SDK code works. - **Keys and billing:** https://console.trymorse.ai. Pay as you go, no subscription; each request costs the model's listed price on https://trymorse.ai/inference/models. - **Privacy per request:** Private (providers under zero data retention agreements), TEE (the model runs on attested hardware its provider can't see into), or E2EE (the request is encrypted on your machine and opened only inside a secure enclave). A request no endpoint can meet is refused, never sent lower. - **Sent as Morse:** providers see Morse as the customer. Your account, card, and IP address stop at Morse. - **Open-source enclave:** https://github.com/trymorse/morse-enclave, checkable with the command above. - **Calls come from your server.** Browsers on other sites can't call the API, so keep your key on a server. - Closed models (GPT, Claude, Gemini), images, web search, and transcription are in the app, not on the API yet. Example: ``` curl https://inference.trymorse.ai/v1/chat/completions \ -H "Authorization: Bearer $MORSE_API_KEY" \ -H "Content-Type: application/json" \ -d '{"model": "moonshotai/kimi-k3.1-pro", "messages": [{"role": "user", "content": "Hello"}]}' ``` Full docs: https://trymorse.ai/inference/docs (and https://trymorse.ai/inference/docs/llms.txt).