TypingMind, ChatWise and Msty Studio all charge for the client and leave the model bill entirely to you, so the decision is not really about price — it is that TypingMind runs in a browser and the other two are native desktop apps. That one difference sets the cheapest entry point, the device limits, whether your chats sync, and whether a third-party endpoint even works: a browser-resident client can only reach an endpoint that sends CORS headers back to it.
Three name collisions to clear first, because all three send searchers to the wrong product. MSTY is also the US ticker of an option-income ETF, and nothing on this page describes that fund. Msty Go — formerly Msty Claw — is a different product from the same company; Kunavo documents it at Msty Go, and its configuration fields are not Msty Studio's. TypingMind is two products under one brand: one-time personal licences at typingmind.com, and a separate per-month subscription called TypingMind Teams at custom.typingmind.com. And the product here is the desktop app distributed from chatwise.app, whose data directory is named app.chatwise — not any similarly named mobile app.
Who should pick which
Pick ChatWise if you work on one or two machines, want the lowest software fee, and want the widest protocol coverage. It is the only one of the three whose custom-provider documentation names an Anthropic-compatible option alongside the OpenAI-compatible one, and it can fetch a provider's model list automatically when the API exposes /models (custom provider docs, September 21, 2026). The cost is portability: its home page says "All your data is stored locally, and never leaves your device (except for sending chat requests to your LLM provider)", and the route it documents between machines is a manual copy — "If you want to move your data to another computer, you can copy the files mentioned above to the same directory on the new computer" — meaning app.db and the files/ directory under %APPDATA%\app.chatwise or $HOME/Library/Application Support/app.chatwise (backup and export docs, September 21, 2026). Whether a hosted sync is planned is not something its documentation states either way.
Pick TypingMind if you want the same history in a browser on any machine, or you want its plugin, Projects and Artifacts workflow. Its cloud sync is opt-in — the install guide tells you to enable it by signing in with your email — and has a published free quota: "By default, you will have 50MB of free cloud storage (typically enough for 5000 chats)", with 1GB and 5GB upgrades whose prices the docs do not print (cloud storage docs). ChatWise documents no comparable hosted sync; its route between machines is the manual file copy above. TypingMind's licence covers 5 devices per key and, at the limit, "our system automatically removes your oldest devices" (manage license and devices). The cost is the browser constraint below, plus tier gating on plugins: the plan matrix in the docs gives Web Search and Dall-E 3 from Extended up and marks "Use unlimited plugins" as Premium only, while the buy page lists plugins under Standard — so check which plugins you actually need before settling on the $39 tier (plan matrix).
Pick Msty Studio if you want a real native app on Linux as well as macOS and Windows — its download page lists macOS 11+, Windows 10/11 x64 and Linux as an AppImage or a .deb — or if you want to try before paying anything: the free tier is $0 forever and its matrix marks Model Hub — Online Models, the bring-your-own-key path, as available on Free. The costs are three. Its terms restrict free use to "your own, private, non-commercial purposes". Aurum caps you at 2 desktop devices. And the same matrix marks Toolbox — MCP Tools "Limited" on Free without ever saying what the limit is, so budget for Aurum if MCP tooling is the reason you are here.
What each vendor charges for the software
| Plan | Price | Shape | Terms worth reading |
|---|---|---|---|
| Msty Studio Free | $0 forever | Free tier | Desktop app itself marked "Limited"; private non-commercial use only |
| ChatWise Free | $0 | Free tier | "Basic chat features"; personal, non-commercial use only |
| ChatWise Pro | $29, with $49 struck through | One-time | Promotional; 7-day trial, "no card required"; one year of pro feature updates |
| TypingMind Standard | $39 | One-time | No strike-through shown; 5 devices per key; 14-day money-back guarantee |
| TypingMind Extended | $79 | One-time | Adds web search, vision and images, image generation, voice output, document upload |
| TypingMind Premium | $99, with $198 struck through | One-time | Promotional; adds multi-model chats, Projects, Artifacts and Canvas, unlimited plugins |
| TypingMind Bulk | $395, with $790 struck through | One-time | Promotional; one shared key for up to 10 users on 50 devices, Premium benefits |
| Msty Studio Aurum | $149 USD per user per year | Auto-renewing | Paid features stop at expiry; 2 desktop devices; unlocks Studio Web |
| Msty Studio Aurum Lifetime | $349 USD per user | One-time | Lifetime updates, protected from future increases, still 2 devices |
Read on typingmind.com/buy, chatwise.app/pricing and msty.ai/pricing on September 21, 2026. Two notes. The struck-through figures are the vendors' own promotional displays and can change without notice, so confirm in checkout. And older Msty reviews quote $129 a year and $249 lifetime because those were the prior figures before an increase the vendor announced for December 2, 2025 — do not budget from them. The Msty device caps, expiry behaviour and non-commercial wording quoted on this page come from msty.ai/terms.
If someone quotes you $83, $166 or $249 a month for "TypingMind", that is TypingMind Teams, a separate subscription product with a Docker and MySQL self-host option and a Business tier the page describes as starting at $18,700 a year. Those three figures are what the page server-renders with its annual-billing toggle showing; the monthly-billing view was not confirmed here, so treat them as billed-annually rates and check the toggle yourself. Separately, TypingMind is also on Setapp at a membership price starting at $14.99 a month — that is a bundle subscription, and which licence tier it grants was not verified.
Three different ways to bring your own endpoint
This is where the comparison stops being about money. All three accept a third-party endpoint, and the mechanisms are not interchangeable.
| Client | Where | What the field takes | Protocols documented | The catch |
|---|---|---|---|---|
| TypingMind | Models → + Add Custom Model | The complete chat-completions URL, not a base URL, plus model ID and context length | OpenAI chat-completions for custom models, plus a Gemini API type; each built-in provider's endpoint can be overridden one at a time under Models → gear icon (Proxy), including the api.anthropic.com/v1/messages default | No dedicated API-key field: the documented place for credentials is a Custom Headers row, and the only worked example in the docs is Gemini's apiKey header. A LocalAI troubleshooting row still says custom models do not support API-key authentication, so confirm before you buy. Requests leave the browser, so the endpoint must send CORS headers |
| ChatWise | Settings → Providers → + | API key plus endpoint settings; models fetched from /models or added by hand with ID, display name, capability flags and context length | "Any OpenAI-compatible or Anthropic-compatible AI service" | The docs give no example URL, so the base-URL convention is unstated; its MCP docs say "OAuth is not yet implemented" and that ChatWise sends to Streamable HTTP servers and gets responses but "does not listen for messages sent by the server at the moment" |
| Msty Studio | Model Hub → Model Providers → Add Provider | "Add your API key (and any required endpoint details), then save" | An "OpenAI-compatible endpoint" option is the only third-party shape the Studio docs describe | Those docs list no field labels for the dialog and give no base-URL example, so whether the URL needs /v1 is undocumented rather than known |
The CORS constraint is the one people discover late, so it is worth stating plainly. TypingMind's own documentation says "Typing Mind will send requests to the local model directly from the browser", its Ollama guide sets OLLAMA_ORIGINS to a wildcard, its LocalAI guide passes --cors true, and the first row of that guide's problem table is CORS issues. An endpoint that returns no CORS headers will fail in TypingMind; ChatWise and Msty Studio are native apps and are not blocked on that account, which is not a promise that everything else about your endpoint will work in them. Self-hosting TypingMind does not remove the requirement — it changes which origin has to be allowed, and the self-host docs add that on a hostname other than localhost you must use HTTPS for all app features to work. See OpenAI-compatible API for what that endpoint shape involves generally.
"Self-host" means two different things in a TypingMind quote, and the cheaper one is not the team one. The personal Static Self-host version is listed as a Self-host Package on all three personal tiers and ships compiled files from the TypingMind/typingmind repository; its own FAQ is explicit that "the full source code is not available for sale as the app is not open-source", that entering the licence key makes the app "connect to TypingMind's license server to verify your license", that "as of now, you cannot customize anything", and that its updates run "delayed from 1 or 2 versions" behind typingmind.com. TypingMind Teams self-host is a separate, separately quoted product: a Docker image with a MySQL 8.x database.
What the model bill looks like
These figures are illustrative token arithmetic, not measured task costs and not a bill ceiling. Assume one person runs 40 chat turns a working day, 22 days a month, for 12 months, and that each turn sends 6,000 uncached input tokens and returns 700 output tokens — a growing thread plus a system prompt. That is 63.4M input and 7.4M output tokens a year. Rates come from the live Kunavo catalog.
| Model | Input / output per 1M | Estimated tokens, year one |
|---|---|---|
| Claude Haiku 4.5 | $0.40 / $2.00 | $40.13 |
| Gemini 3.8 Flash | $0.525 / $2.625 | $52.67 |
| GPT-5.6 Terra | $0.70 / $4.20 | $75.40 |
| Claude Sonnet 4.6 | $1.20 / $6.00 | $120.38 |
| Client plan | Software, year one | Plus tokens at Claude Sonnet 4.6 | Illustrative year-one total |
|---|---|---|---|
| Msty Studio Free | $0.00 | $120.38 | $120.38 |
| ChatWise Pro, one-time | $29.00 | $120.38 | $149.38 |
| TypingMind Standard, one-time | $39.00 | $120.38 | $159.38 |
| TypingMind Premium, one-time at the promotional price | $99.00 | $120.38 | $219.38 |
| Msty Studio Aurum, per user per year | $149.00 | $120.38 | $269.38 |
| Msty Studio Aurum Lifetime, one-time | $349.00 | $120.38 | $469.38 |
The reading that matters: under these assumptions the token estimate of $120.38 is larger than four of the six software lines, so picking a client on its sticker price optimises the smaller number. Moving the same traffic from Claude Sonnet 4.6 to Claude Haiku 4.5 changes the estimate by $80.26 — more than the ChatWise Pro and TypingMind Standard licences cost put together. Year two flips the software column again: the one-time licences drop to $0 while Aurum renews. What happens at that point differs, though. ChatWise sells "one year of pro feature updates" on a licence its pricing page describes as "pay once and use it forever", so what lapses is the update window rather than the licence; Msty's terms are explicit that "once the license expires, you will lose access to the paid features". AI cost optimization covers the model-choice half properly.
Kunavo serves no text-to-speech, speech-to-text or embedding model, so TypingMind's Whisper transcription proxy default and the text-to-speech feature on its Extended and Premium tiers cannot be pointed at a Kunavo key, and neither can any transcription or embedding step in ChatWise or Msty Studio — those calls need a provider that does serve those endpoints.
Kunavo's catalog amount is a billing floor rather than a cap: when the upstream reports its charge, the bill is the greater of catalog cost and upstream cost times the applicable markup. Cache charges and external tools sit outside these examples. The minimum top-up is $10 in prepaid credit, which is a funding minimum, not a task fee or a subscription — see billing details.
Setting one up
Kunavo publishes a configuration reference for one of the three. That is a written setup guide, not a compatibility test: none of TypingMind, ChatWise or Msty Studio has been runtime tested against Kunavo's endpoint, and every behaviour above comes from the vendors' own documentation rather than from an observed call. Keep a working route available while you try it, run one bounded task, then read what your account actually recorded. Start at the ChatWise integration guide, and create a Kunavo account when you are ready to fund a key.
Still narrowing the field? ChatWise vs Chatbox covers the free-client alternative, TypingMind vs ChatGPT covers the licence-versus-subscription question directly, and best LLM gateway compares what sits behind whichever client you choose.
FAQ
TypingMind vs ChatWise: which should I buy?
Buy ChatWise if you want a native desktop app and the lowest software fee: Pro is a one-time $29 with $49 struck through, it accepts OpenAI-compatible and Anthropic-compatible custom providers, and it can fetch a provider's model list from a /models endpoint. Buy TypingMind if you want the same chat history in a browser on every machine, if you need its Artifacts and Projects workflow, or if you want hosted sync at all — ChatWise's home page says all your data is stored locally and never leaves your device, and the route between machines its documentation gives is copying the data directory by hand. The constraint that decides it for a self-hosted or gateway endpoint is architectural: TypingMind sends model requests from the browser, so the endpoint must return permissive CORS headers, while ChatWise is a desktop app and never asks the question. Prices and quotes read September 21, 2026.
Msty vs TypingMind: what is the real difference?
Msty Studio is a native desktop app for macOS, Windows and Linux with a free tier at $0 forever, and its own feature matrix marks the desktop app itself as Limited on that free tier without ever saying what the limit is. TypingMind is a browser app — the only install path its documentation offers is a Progressive Web App, with no app store and no download — sold as a one-time licence starting at $39. On money, Msty Studio's paid tier is a recurring $149 USD per user per year, or $349 USD once for Aurum Lifetime, while TypingMind Premium is $99 once at its current promotional price. On reach, Msty caps an Aurum licence at 2 desktop devices and TypingMind allows 5 devices per key, evicting your oldest device automatically when you hit the limit. Both bill model tokens separately. Checked September 21, 2026.
Do TypingMind, ChatWise and Msty Studio include model usage?
No. All three sell software and leave the model bill with you. TypingMind's purchase page carries the footnote that the price does not include API costs and additional cloud storage. ChatWise tells you to use your own API keys. Msty Studio describes its online providers as letting you use hosted models with your API keys without running inference on your device. So the total is always two numbers: a licence or subscription paid to the client vendor, and metered tokens paid to whichever API you point it at. Read on the three vendors' own pages, September 21, 2026.
Can one OpenAI-compatible API key drive all three clients?
In principle yes, because all three accept an OpenAI-compatible endpoint, but the three setup flows are not equivalent and one of them has a hard prerequisite. TypingMind's Add Custom Model form takes the complete chat-completions URL rather than a base URL — its Ollama guide's worked example is http://localhost:11434/v1/chat/completions — and has no dedicated API-key field: the documented place for credentials is a Custom Headers row, whose only worked example in the docs is Gemini's apiKey header, while a LocalAI troubleshooting row still says custom models do not support API-key authentication. Its requests also leave your browser, so the endpoint must serve CORS to the TypingMind origin. ChatWise takes an API key plus endpoint settings in Settings then Providers, and accepts OpenAI-compatible or Anthropic-compatible services. Msty Studio takes an API key and any required endpoint details under Model Hub then Model Providers, and the only third-party shape its docs describe is an OpenAI-compatible endpoint. Documented behaviour read September 21, 2026; none of the three was tested against Kunavo.
Which of the three is cheapest?
On the software line alone, Msty Studio Free at $0 forever, then ChatWise Pro at a one-time $29, then TypingMind Standard at $39. But the two free-or-cheap options carry conditions the price tag does not show: Msty Studio's terms restrict free use to private, non-commercial purposes and say using it for the trade or profession you are compensated for does not qualify, and ChatWise's pricing page likewise marks its free tier "Free for personal, non-commercial use" without defining the boundary on the page itself. For paid professional work the honest comparison is ChatWise Pro $29 once against TypingMind Standard $39 once against Msty Studio Aurum $149 USD per user per year. Beyond that, the cheapest listed software price and the lowest total cost of your work are different questions, because the token bill is usually the larger of the two numbers.
Is TypingMind a desktop app?
Not in the usual sense. TypingMind's own install documentation says it runs at typingmind.com in any modern browser and that you can also install it as a Progressive Web App on any platform, with no app store and no download required. That matters beyond packaging: because the page itself issues the model requests, a custom or self-hosted endpoint has to return CORS headers that allow the TypingMind origin. TypingMind's Ollama guide sets OLLAMA_ORIGINS to a wildcard and its LocalAI guide passes a CORS flag, which is what that requirement looks like in practice. ChatWise and Msty Studio are native desktop applications and do not impose it. Documentation checked September 21, 2026.
Vendor pricing pages, licence terms and setup documentation for all three products checked September 21, 2026; TypingMind Teams pricing, Setapp membership pricing and the Msty price-increase announcement were read the same day. Kunavo token rates come from the live catalog, and every dollar example here is illustrative token arithmetic rather than a measured bill. No runtime test of any of the three clients against Kunavo was performed.