TypingMind pricing: the licence, Teams, and the API bill behind both

TypingMind pricing is two separate price lists, and neither one includes model tokens. The individual licence is one-time — $39 Standard, $79 Extended, $99 Premium shown as 50% off $198, plus a $395 bulk key — while TypingMind Teams is a subscription that the pricing page shows at $83, $166 and $249 per month in its annual-billing view. The third line item is the one people miss: TypingMind's own FAQ says "TypingMind does not sell API credits", so the model bill is yours, it is metered, and it does not stop when a "lifetime" licence is paid.

Three URLs look like a price page and only one is. typingmind.com/buy carries the real individual list; typingmind.com/pricing redirects to a homepage anchor with no table; and the docs licence plans page names tiers without printing amounts. Separately, custom.typingmind.com is a different product — TypingMind Teams — and most third-party "TypingMind pricing" articles blend the two lists into one. All checked September 21, 2026.

What the licence costs

Read from the buy page and the general FAQ on September 21, 2026. The buy page states "One-time, no renewal", a 14-day money-back guarantee, upgrades between tiers at any time, and an offline static self-host build included with the licence. Two cautions: the Premium and Bulk struck-through prices are a promotion with no published end date, so treat $99 as today's price rather than a permanent one; and the licence's reach is stated two ways — "Each license key works on up to 5 devices" in the FAQ against "5 devices or users per license key" in the Personal-versus-Teams page. Also note the GitHub repository is a distribution of the compiled self-host build, not an open-source project: its LICENSE.md opens "NOTE: This is NOT an open-source software."

That buy-page footnote — the price "does not include API costs and additional cloud storage" — names both recurring lines behind a one-time purchase. The model bill is the large one and it never stops. The other is TypingMind Cloud sync: the cloud-storage documentation gives 50MB free, "typically enough for 5000 chats", then offers 1GB and 5GB plans you can "pay monthly or annually" — and prints no figure for either, since those are visible only inside the app. The amounts circulating on third-party review sites do not come from TypingMind, so this page will not repeat them.

TypingMind Teams is the other price list

The annual figures and the Business and self-host rows were read directly from the Teams pricing page on September 21, 2026, which renders the annual view by default and labels it "2 MONTHS FREE". The monthly column is not the one that page shows by default; those figures come from TypingMind's plan comparison docs, which print Starter $99, Growth $199 and Professional $299 per month with no struck-through list price at all, and which also state seats for every plan: "All plans include 5 user seats. Additional seats can be added at $8/user/month." The two sources do not contradict each other on rates, but they lead with different numbers — $83 a month on the pricing page against "starting at $99/month" in the docs — which is worth knowing before you quote one to a finance team. Neither prints an annual total, only a per-month equivalent, so confirm the figure that will actually be charged in checkout. The pricing page carries one notice under its plan grid, repeated under the self-host block: "API cost not included, you need to pay OpenAI, Anthropic, or Google for your API usage separately".

One Teams detail genuinely changes the architecture. Its pricing FAQ says "Your users will not see your API key. Requests to the OpenAI API are sent from our server." That statement names the OpenAI path specifically; TypingMind does not extend it to models you add yourself, so do not assume a custom endpoint on Teams is called server-side. One name trap goes with it: the API Access feature is "A REST API over JSON for your chat instance" for managing users and agents, Professional tier only in the plan comparison. It administers your instance; it does not sell tokens.

Best API for TypingMind: the browser filters the field before price does

The product this page is about runs in a browser. TypingMind's own install page offers only a Progressive Web App — "No app store, no download required" — and its FAQ says "All API requests go directly from your browser to the AI provider's servers". So every model call is a cross-origin request from the typingmind.com origin, or from whatever origin your static self-host copy is served from, and the endpoint must return CORS headers permitting that origin or the browser rejects the call at preflight before the provider sees it. Self-hosting the static build does not change this, because the request still starts in a browser. TypingMind's Local AI page says the same thing from the other side: its worked command adds --cors true "to make sure the local server is accessible from the browser", and its troubleshooting table lists CORS as the first problem to expect. One exception this page did not test: TypingMind also ships a macOS app through Setapp, whose own FAQ says the Setapp free models are "only accessible via the macOS app in Setapp" — nothing below was checked against that build.

Those five preflights were run from a terminal for this page at 14:20 UTC on September 21, 2026. In the same day's run, Vercel AI Gateway, Gemini's OpenAI-compatible path, DeepSeek, Groq and MiniMax also returned headers permitting the origin. So Kunavo cannot be reached from TypingMind's browser app today, and this page will not pretend otherwise; the Z.ai row is a caution rather than a verdict, since Z.ai appears in TypingMind's built-in provider list and its integration may use a different path than the public one tested here.

Two configuration routes exist once an endpoint qualifies. The Proxy setting overrides a built-in provider's endpoint — its documented defaults are https://api.openai.com/v1/chat/completions, https://api.anthropic.com/v1/messages, https://generativelanguage.googleapis.com/v1beta/models and a Whisper transcription URL that Kunavo does not serve — and it is set per provider, from the gear icon beside that provider in Models. Add Custom Models is the other: TypingMind's worked example shows a complete chat/completions route rather than a base URL, a hand-typed model id and context length, a Test button, a manual Support Plugins toggle, and no key field at all — the key goes in as a custom header named Authorization with the value Bearer your_api_key. There is no single "add a custom model" page in the documentation; that pattern is repeated per provider.

Best model for TypingMind: the official page is a wall, and a dated one

TypingMind's "Which AI Models Should You Use" page is where this question lands, and on September 21, 2026 it still opened with "Choosing the right AI model in 2025" and carried an API Pricing row with a single value in it — "Free to use" under GPT — leaving Claude, Gemini, Perplexity, Grok, Llama and DeepSeek blank. Since the app is bring-your-own-key, the answer is your provider's current list price, filtered by the CORS condition above.

Rates from OpenAI, Google and Anthropic, standard processing tiers. The Gemini row is the one that rots: its published promotional price has a stated expiry, so a budget built on $0.75 changes in January. And treat TypingMind's own cost estimate as a rough order of magnitude rather than a budget — the general FAQ says a 1,000-character conversation "costs roughly $0.001", but attaches no model and no date, so there is no rate above it can be checked against.

A worked estimate: licence plus tokens against a monthly plan

TypingMind's buy page frames its own pitch as "$39 once + pay per API use" against "$20/mo · $240/yr forever", so that is the comparison worth doing with real arithmetic. This is illustrative token arithmetic, not a measured bill and not a ceiling. Assume 120 chat turns a month, each sending 6,000 uncached input tokens as the conversation grows and returning 700 output tokens: 720,000 input and 84,000 output tokens a month. The first-year column adds a $99 Premium licence once. Rates are live Kunavo catalog prices per million tokens.

The shape is the finding, not the cents. At this volume every row finishes its first year under the $240 a year TypingMind names for a subscription, and the spread between the cheapest and the dearest model is about $26 across twelve months — less than the $60 between the Standard and Premium tiers. In other words the licence decision outweighs the model decision for a light chat habit, and "metered or flat" is a question about volume rather than about models. The break-even against that anchor is roughly $11.75 of tokens a month once the licence is paid, which is about 620 turns on Claude Opus 5 and far more on GPT-5.6 Luna. Raise the assumption tenfold, to 1,200 turns a month, and the picture changes: Claude Opus 5 lands near $373 for a first year, past that anchor, while GPT-5.6 Luna is still near $109. Scale by your own turn count before treating any row as a budget, and note that these rates belong to a client that can reach the endpoint — for TypingMind itself today, apply the provider rates in the previous table instead.

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 this example. The minimum top-up is $10 in prepaid credit — a funding minimum, not a task fee or a subscription. See billing details.

Which route wins

Direct vendor keys win when you live inside one vendor's models: OpenAI and Anthropic both permitted the browser origin in the preflights above, Anthropic only once the request declares a header it names dangerous, which tells you how it regards keys held in a web page. A CORS-permitting gateway wins when you switch models per task and want one balance rather than several. A subscription wins when heavy daily use suits a flat fee better than a meter — that is the case TypingMind argues against on its own buy page, and it is a real case. A local model wins for private or small work at no per-request charge, provided you start it in CORS mode. One boundary applies to every route: Kunavo serves no embedding, text-to-speech or speech-to-text model, so TypingMind's voice and retrieval features stay with the client or with another provider whichever chat endpoint you choose.

If the route you want is a gateway whose endpoint does not permit the browser origin, the fix is not a setting — it is a client that issues the call from its own process instead of from a web page. A native desktop app, or a server that calls the provider on your behalf, is not making a browser cross-origin request, so the preflight above never happens. TypingMind alternatives compares nine of them on licence and price; Kunavo publishes setup pages for several, including Cherry Studio and Open WebUI. Those are configuration references, not compatibility tests — no client here has been runtime-tested against Kunavo's endpoint — so keep a working provider available while you try one, run a bounded session, and read the charge your account records. Create a Kunavo account when you want to fund that route. For rate comparisons by family, see Claude API pricing, GPT API pricing and Gemini API pricing, and AI cost optimization for cutting the token bill itself.

FAQ

Read directly while writing this page on September 21, 2026: TypingMind's buy page, its general FAQ, the Personal-versus-Teams page, the Teams pricing page, the Teams plan comparison, the Teams API Access page and the "Which AI Models Should You Use" page — plus the CORS preflights in the table above and the five further endpoints named beneath it, all run from a terminal at 14:20 UTC that day. Checked the same day in this page's source pass and not re-opened here: the licence-plans page, the cloud-storage page, the install page, the Setapp FAQ, Proxy settings, the MiniMax custom-model example, the Local AI page, the repository LICENSE.md, and the published rate cards of OpenAI, Google and Anthropic. Not checked at all: TypingMind's unpublished cloud-storage prices, the annual total a Teams subscription actually charges at checkout, when the current promotions end, whether Teams routes custom-model calls server-side, how the Setapp macOS build reaches a provider, and whether any of this works end to end inside TypingMind — nothing here was runtime-tested in the app, and no client on this page has been runtime-tested against Kunavo. Kunavo token rates come from the live catalog and every dollar example is illustrative token arithmetic.