Morse vs Kagi: where your AI threads are kept, who can read them, and price
The Morse teamPublished Facts checked 10 min read
Morse and Kagi both keep ads and AI training out of your chats, as of November 2026, and they split on where the conversation sits afterwards. Kagi Assistant saves each thread in Kagi's database under your Kagi account, kept 24 hours after the last message by default or longer if you choose, while Morse stores your chat history in your browser, encrypted, not on Morse's servers. Choose Kagi if you want a paid search engine with AI on top; choose Morse if you want 80+ models, open and closed, with no thread on Morse's servers and end-to-end encrypted chats your own browser checks.
We read Kagi's pricing page, its privacy policy (last changed September 22, 2026), its Assistant and LLMs & Privacy help pages, its plan types and Privacy Pass docs, its AI philosophy page, its changelog, and Kagi's own answer on its feedback forum about how threads are stored, all on November 18, 2026. Every Kagi fact links to the page it came from. To Kagi's credit, it "is fully funded by its users" and shows no ads, its privacy policy says it doesn't "load any analytics or telemetry," and it publishes how long each model provider keeps an Assistant request (Kagi, Kagi).
Morse vs Kagi at a glance
As of November 18, 2026.
| Morse | Kagi Assistant | |
|---|---|---|
| Where your prompt is decrypted | Anonymous: at the provider, without knowing it's you. Private: at a host that agreed to keep nothing. E2EE: only inside an enclave your browser verified | On Kagi's servers, which proxy "all connections and data" for the Assistant (Kagi), then at the model's provider |
| Who verifies it | E2EE: your browser checks the enclave's attestation before sending. Private: Morse checks each reply where a provider confirms zero retention on it | Kagi's word and its providers' policies; its own table marks one model "Not verified" (Kagi) |
| Where history lives | In your browser, encrypted, not on Morse's servers | In Kagi's database under your account: 24 hours after the last message by default; 7 days, 30 days, or permanent if you choose (Kagi, Kagi) |
| End-to-end encryption | E2EE chats on Pro | None described; threads are "available as plaintext" to the Assistant application and "operations engineers with monitored access" (Kagi Feedback) |
| Training | Morse doesn't train AI on your chats | "Not by us or by the LLM providers" (Kagi) |
| Ads | None on any plan | None |
| Models | 80+, open and closed; uncensored ones on Pro and Max | 34 in the Assistant's list: 12 on every plan, 22 on Ultimate only |
| When you hit the limit | Free: 15 messages a day. Pro: unlimited on Pro models, plus 250 credits for top models | AI "will be disabled" until the next billing cycle, unless you upgrade or renew early (Kagi) |
| Web search | On the Free plan; none in E2EE chats | Kagi Search, with your Lenses and personalized results |
| Apps | Web app | Web, plus native iOS and Android apps (Kagi) |
| Legal requests | We usually have very little to give authorities, because we don't store chats | A warrant canary listing "0 Warrants from any government organization" (Kagi) |
| Price | Free (15 messages a day); Pro $19 a month; Max $49 | Trial (100 Assistant interactions); Starter $5; Professional $10; Ultimate $25 |
Is Kagi Assistant more private than Morse?
Not for your history, and about even for what the model's maker sees. Kagi says "interactions with the Assistant are stored in threads," kept "for 24 hours after the last message" unless you change Thread Saving, and that the setting "is applied when the thread is created" (Kagi). Its August 21, 2026 changelog added 7-day and 30-day options beside 24 hours (Kagi).
Who on Kagi's side can open a saved thread, what a reported thread leaves behind, and why Privacy Pass doesn't cover the Assistant yet are in is Kagi Assistant private.
What does each model's provider keep, in Kagi and in Morse?
The same model family can land in a different place on each service. Kagi's LLMs & Privacy table gives a retention period for each provider, and its Assistant page gives the plan. Here they are beside the Morse mode that runs each family, read November 18, 2026.
| Model family | What the provider keeps on Kagi | Kagi plan | Mode on Morse | What the host keeps on Morse |
|---|---|---|---|---|
| Claude | 30 days; flagged prompts 2 years | Haiku on every plan; Sonnet, Opus, and Fable on Ultimate | Anonymous | The provider sees the message without knowing it's you and may keep it for a limited time to check for abuse, under its own policy |
| GPT | 30 days | GPT 6 Luna, Astra, Sol, and Terra on Ultimate; GPT OSS 120B, an open-weight model marked "Not stored," is on every plan | Anonymous, for GPT-6 models | Same as above |
| Gemini | 24 hours; Gemini 4 Carbon up to 30 days through Google Vertex | Ultimate, except Gemini 3.1 Flash Lite | Anonymous | Same as above |
| Grok | 30 days | Ultimate | Private | Hosts agreed to keep nothing |
| DeepSeek, Kimi, and GLM | "Not stored" | DeepSeek V4.1 Flash and GLM 5.3 Flash on every plan; the rest on Ultimate | Private, or E2EE on Pro | Private: hosts agreed to keep nothing. E2EE: only an enclave your browser verified can read it |
Will Kagi Assistant or Morse refuse a question?
Kagi answers with each provider's own model, and we found no refusal rules of Kagi's own on the pages we read. None of the 34 models in its list is labeled uncensored (Kagi). Kagi frames AI as a search tool first. Its AI philosophy page says AI "should be used in a closed, defined context relevant to search (don't make a therapist inside the search engine, for example)," and that it should "enhance the search experience, not to create it or replace it" (Kagi). Its report form lists "potentially prohibited content" among the reasons to flag a reply (Kagi).
On Morse's Pro and Max plans, picking an uncensored model such as DeepSeek V4.1 Flash Uncensored gives answers other apps refuse: dark fiction, straight medical and legal answers, and help testing your own security. Uncensored here means answers instead of refusals on health, law, money, security, and politics. The list of uncensored models changes, so check the app's model menu.
Which has more models, Kagi or Morse?
Morse, with 80+, open and closed, from OpenAI, Anthropic, xAI, DeepSeek, Z.ai, Moonshot AI, Qwen, and others. Kagi's Assistant list names 34 models from 12 developers, from Alibaba to Z.ai (Kagi). Twelve are open to every plan: Claude 5.5 Haiku, DeepSeek V4.1 Flash, GLM 5.3 Flash, Gemini 3.1 Flash Lite, GPT OSS 120B, three Mistral models, two Qwen models, MiniMax M3.1 Flash, and Hermes 4 405B. Every other Claude, GPT, Gemini, and Grok model needs Ultimate.
Kagi's own pages don't agree on which versions Ultimate gets. On November 18, 2026, its pricing page lists Claude Fable 5.1, GPT-6 Astra, GPT-6.1 Sol, GPT-5.6 Terra, and Kimi K3 (Kagi). The Assistant's model list names Claude Fable 5.2, GPT 6.2 Astra, GPT 6.2 Sol, GPT 6.1 Terra, and Kimi K3.1 Pro. The same help page's opening line names Amazon among its model makers, though no Amazon model appears in the list. If you're paying $25 for one particular model, check Kagi's model menu during the trial first.
Is Kagi cheaper than Morse?
Kagi is cheaper to start, and Morse is cheaper for Claude Sonnet and Grok. Kagi's prices also buy its search engine. Morse is a chat app, and web search comes with its Free plan.
| As of November 18, 2026 | Price | What the AI side includes |
|---|---|---|
| Kagi Trial | Free | 100 searches and 100 Assistant interactions (Kagi) |
| Kagi Starter | $5 a month | 300 searches, and "up to 300 AI interactions per month with standard models" (Kagi) |
| Kagi Professional | $10 a month | Unlimited searches, and the Assistant's Quick mode with "a larger monthly usage allowance" (Kagi) |
| Kagi Ultimate | $25 a month | Research mode and flagship models, with "$25 worth of token usage" a month, or $270 a year on yearly billing (Kagi) |
| Morse Free | $0, no card | 15 messages a day, with the model picked for you |
| Morse Pro | $19 a month, or $17 billed yearly | Unlimited messages on Pro models such as Claude Sonnet 5.5, Grok 4.7, and DeepSeek V4.1 Flash Uncensored, plus 250 credits for top models such as Claude Fable 5.1 and GPT-6 Astra; E2EE chats; encrypted sync across 3 devices |
| Morse Max | $49 a month, or $44 billed yearly | Everything in Pro, plus 5,500 credits worth $55 and encrypted sync across 5 devices |
Kagi's $25 is a budget, and part of it isn't tokens. Its help page says "$25 token usage consists of $20 for raw token costs and $5 for operational costs," which covers search queries, infrastructure, and development (Kagi). That $5 is 20% of the plan price, or a 25% markup on the $20 of raw token cost. Claude Sonnet 5.5, Grok 4.7, and every other Ultimate model draw on that same $20, and when it runs out, "new AI interactions will be disabled" until the next billing cycle unless you renew early. On Morse Pro, messages on Pro models such as Claude Sonnet 5.5 and Grok 4.7 are unlimited, and credits are only for the top models: 100 credits cost $1, so Pro's 250 are worth $2.50 and Max's 5,500 are worth $55. Both companies take 10% off for yearly billing.
Where Kagi is better
- The search engine. The Assistant's web access is "powered by Kagi Search," and you can apply your Lenses and Personalized Results to its searches (Kagi). Kagi's FAQ says "queries you type are never associated with your account" (Kagi), and Privacy Pass can unlink searches from your login on Professional and up (Kagi).
- Phone apps. Kagi Assistant has had native iOS and Android apps since July 30, 2026, and "your threads and Custom Assistants stay with you" across them (Kagi). Morse is a web app.
- Files, voice, and custom assistants. The Assistant takes uploads up to 30 MB, including PDFs, spreadsheets, images, and audio, fetches web pages up to 50 MB, and has voice input and custom assistants with their own model and instructions (Kagi).
- Paying only for months you use. In a month you don't use your plan, Kagi says "we will automatically apply a full credit to your account" (Kagi).
- A published canary. Kagi's privacy policy lists "0 National Security letters," "0 Gag orders," and "0 Warrants from any government organization" (Kagi).
How Morse keeps your chat history in your browser
A Kagi Assistant thread is a row in a hosted database under your Kagi account, and that row lasts as long as your Thread Saving choice says. Morse is built so that row doesn't exist. Your chat history is stored in your browser, encrypted, and Morse's servers don't keep messages or log what you write. There's no thread timer to pick before you start, a temporary chat is kept in memory only, and Morse doesn't train AI on your chats, sell your data, or show ads (Morse privacy).
What a Kagi user gets after switching, by the models they used:
- Kagi's every-plan DeepSeek and GLM, now in Private. Kimi runs here too, and Grok 4.7 comes with Pro; both needed Ultimate on Kagi. New chats start in Private. 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.
- Claude, GPT, and Gemini, now in Anonymous. Morse's proxy strips your account, email, and IP address and sends under Morse's own account. The provider sees the message without knowing it's you and may keep it for a limited time to check for abuse, under its own policy, so leave your name out.
- The thread you'd rather not leave in a hosted database, now in E2EE on Pro. Your browser checks the enclave's attestation and seals the message for it, so Morse's servers relay text they can't read. It runs open-weight models with no web search or image generation, and rests on the enclave hardware and attestation working as designed (how the check works).
Because we don't store chats, we usually have very little to give authorities. Clearing your browser, losing the device, or deleting your account deletes the chats on it, and Morse can't recover them, so export what you need from Kagi first and copy out any Morse answer you'll want again. You sign in with a code sent to your email, or with Google.
Should you choose Kagi or Morse?
- Choose Kagi if search comes first: you want Lenses and Personalized Results shaping the AI's answers, native phone apps, 30 MB uploads with audio, a $5 starting price, a month's credit when you don't use it, and a published warrant canary.
- Choose Morse if the models come first: 80+ instead of 34, Claude Sonnet 5.5 and Grok 4.7 as unlimited messages on Pro at $19 rather than a share of Ultimate's $20 token budget, uncensored models on Pro and Max for questions other apps refuse, and chat history stored in your browser, encrypted.
Other private options sit side by side in Kagi Assistant alternatives, and Kagi's search against Perplexity's is in Kagi vs Perplexity. Try a question on the Morse home page with no account, or sign up free and move to Pro when you want E2EE.
Morse publishes this comparison. We've tried to keep it fair; if something's wrong or out of date, email hello@trymorse.ai.
Kagi is a trademark of Kagi Inc. Morse is not affiliated with or endorsed by Kagi Inc.
Common questions
Is Kagi Assistant end-to-end encrypted?
Kagi's pages don't describe it that way, and Kagi has said saved threads are available as plaintext to the Assistant application and to operations engineers with monitored access. On Morse, E2EE chats on Pro are sealed on your device for one enclave your browser has verified. The guarantee rests on the enclave hardware and attestation working as designed.
Is Morse cheaper than Kagi?
It depends on the models you want. Kagi Starter, at $5 a month, is the cheapest paid plan of the two, and Morse's Free plan costs nothing for 15 messages a day. For Claude Sonnet 5.5 and Grok 4.7, Kagi asks for Ultimate at $25 a month, and Morse Pro gives unlimited messages on both for $19.
Can I use Kagi Search and Morse together?
Yes. They're separate services with separate accounts. Kagi says the queries you type into Kagi Search are never associated with your account, so you can keep searching on Kagi and hold your AI chats in Morse, where history is stored in your browser.
Does Kagi train AI on your Assistant chats?
No, according to Kagi's Assistant page, not by Kagi and not by the model providers. Morse doesn't train AI on your chats on any plan either.