dalao.ai

TypingMind

A better UI for your own AI keys

About TypingMind

TypingMind is an AI chat front end for people who bring their own API keys: models from OpenAI, Anthropic, Google and others in one browser interface, with chats kept in the local browser by default. Personal licenses are one-time: $39, $79 or $99 as of 2026-10-25; TypingMind Teams is a hosted subscription from $83 a month in the annual view. No tier includes model usage, which is paid directly to the model providers.

EditorialUpdated 10 min read

What it is

TypingMind was launched by Tony Dinh in March 2023, starting as a web interface for the ChatGPT API. Users enter API keys from OpenAI, Anthropic, Google Gemini and other providers; TypingMind supplies the interface, conversation management, plugins and agents. By default the browser sends model requests straight to the provider, and usage is billed by that provider.

There are two product lines. The personal edition (the TypingMind License) is a one-time purchase that works on the web, as an installable PWA, or as a downloadable static self-host package. TypingMind Teams is a hosted subscription with an admin panel, preset API keys, knowledge bases, member permissions and SSO. The docs list four forms: the personal license, the free static self-host, Teams, and Teams self-host on a custom quote.

Data stays on the device by default. The Privacy Policy (updated 2026-09-19) says that in local mode chats sit in the browser's LocalStorage and IndexedDB with no backend reading them; self-hosted deployments send no data to TypingMind; cloud sync is opt-in. There is no native iOS or Android app; phones use the PWA.

The Terms of Service (updated 2026-09-17) name Typing Technology Co. Ltd. as the contracting party, governed by the laws of Singapore, subject to mandatory consumer protection law.

All your data remains securely stored on your device.

TypingMind homepage, Privacy-first local storage panel

Core capabilities

Many models side by side. The homepage lists OpenAI, Claude, Gemini and more than 20 other model sources; the Premium tier lets several models answer the same prompt in parallel. The pricing page's comparison table sums up coverage as "OpenAI, Claude, Gemini + 20 more".

Agents, plugins and knowledge sources. The homepage claims more than 50 pre-built AI agents, plus MCP servers, plugins and RAG sources (GitHub, Notion, Google Drive), deep research, Artifacts and a Canvas editor. Premium removes the cap on plugins and MCPs; Standard already includes agents and plugin sharing.

Cost control. The homepage says prompt caching can cut token costs by up to 90%, as reported by the vendor. The changelog entry of 2026-10-21 adds a 12-hour OpenAI prompt caching option.

Release pace. In the October 2026 changelog, memory controls arrived on 2026-10-14, DeepSeek V4.1 Flash on 2026-10-19 and the 12-hour OpenAI prompt caching option on 2026-10-21. The pricing page claims more than 100 updates in the last 6 months, as reported by the vendor.

Team features. From the Professional tier up, Teams offers an analytics dashboard, audit logs, chat log recording, role-based access, usage limits per user, model or agent, and SSO, Directory Sync and OAuth 2. The Teams homepage says TypingMind is SOC 2 Type II certified and GDPR compliant, as reported by the vendor. Teams can also choose the EU data center at signup, with both the web server and the database in the EU.

Who it's for / Not for

Good fit:

  • Individuals already paying for several model APIs who want one interface, especially light users for whom pay-per-use beats a fixed monthly plan. The pricing page sets "$39 once + pay per API use" against ChatGPT Plus and Claude Pro at "$20/mo"; that comparison is the vendor's own framing.
  • People who want to pay once: the personal license never renews, is valid indefinitely and can be refunded within 14 days.
  • Small teams that want a shared AI front end without maintaining an open-source project: Teams has an admin panel, permissions and SSO, with a 14-day trial.
  • People who want data to stay on the device: by default chats live only in the browser and requests go straight to the provider.

Poor fit:

  • Anyone who needs to audit or modify the source: the static self-host package ships compiled code only, and the terms forbid modification, reverse engineering and redistribution.
  • Anyone without API keys who does not want to manage usage billing: no tier includes model costs, and accounts must be opened with each provider.
  • Anyone who needs a native mobile app: only a PWA is offered.
  • Anyone who needs end-to-end encrypted sync: the cloud sync docs describe encryption in transit and AWS encryption at rest, not end-to-end encryption.

How it compares

LibreChat is an MIT-licensed open-source chat front end; its GitHub page showed 46,279 stars as of 2026-10-25. It can be fully self-hosted, modified and run with multi-user login, but the database, upgrades and security patches are the operator's job.

Open WebUI showed 156,689 GitHub stars as of 2026-10-25. Its license is BSD 3-Clause plus a branding clause: the "Open WebUI" branding may not be removed or altered unless end users stay at 50 or fewer in any rolling 30-day period, prior written permission is obtained, or another listed exception applies.

LobeHub combines an open-source project with a hosted subscription. On its pricing page as of 2026-10-25, Starter is $12.9 billed monthly or $118.8 billed yearly (shown as $9.9 a month in the default view) with 5,000,000 credits a month; Premium is $24.9 monthly or $238.8 yearly. The hosted plans convert model costs into credits.

ChatWise is a desktop chat client that keeps data locally. On its pricing page as of 2026-10-25 there is a free tier for personal non-commercial use, and Pro is $49 one-time with one year of Pro feature updates and a 7-day trial.

Getting started

  1. Open typingmind.com and try it in the browser. According to the pricing FAQ, the free plan does not save chat history, shows a popup every few minutes and limits some features.
  2. Enter model provider API keys in settings. The pricing FAQ says keys are stored only locally and encrypted with AES; if cloud sync is turned on later, keys sync along with other data.
  3. After buying, activate with the license key. Each license covers up to 5 devices; when the limit is hit, signing in on a new device removes the oldest one automatically.
  4. Cloud sync lets several devices share the same data. The free allowance is 50MB, which the vendor estimates at about 5,000 chats; paid upgrades exist.
  5. For offline or intranet use, deploy the static self-host package following the GitHub repository TypingMind/typingmind. It is a static front end, and chat data still stays in the local browser.
  6. Team trials start at custom.typingmind.com with a 14-day free trial; teams that need EU data residency must pick the EU data center at signup.

Pricing & license

Personal licenses (pricing page, 2026-10-25; all one-time):

  • Standard $39: no ads, basic chat, agents, chat and plugin sharing, voice input.
  • Extended $79: adds web search, vision, image generation, text-to-speech and document upload.
  • Premium $99: shown as half off a $198 list price with a countdown; adds parallel multi-model chats, project folders, Artifacts and Canvas, unlimited plugins and MCPs, and lists "Free updates".
  • Bulk license: 10 users and 50 devices for $395 (struck-through $790).

These prices exclude API costs and extra cloud storage, carry a 14-day refund window, and can be upgraded at any time by paying the difference. The FAQ says the Premium license is valid forever and includes all future updates; for Standard and Extended it only says updates depend on the license bought.

Teams (pricing page, 2026-10-25, annual billing selected by default): Starter $83 a month with 5 seats and $8 per extra seat per month; Growth $166 a month; Professional $249 a month (struck-through $499); Business on quote, starting at $18,700 a year and as low as $4 per seat per month. Team licenses auto-renew each term and are non-refundable under the terms, except where consumer law says otherwise.

Price history: the founder's newsletter of 2023-03-10 describes an early-bird price of $19, then $29, then $39; the "Launched a new pricing!" entry in the docs changelog links a tweet posted at 2023-05-14 16:32 (UTC+8).

License: the personal edition is proprietary. LICENSE.md in the static self-host repository was first added by commit a619c1e ("Init", author Tony Dinh) at 2024-02-09 22:57 (UTC+8). It opens with "This is NOT an open-source software", ships compiled code only, and forbids modification, redistribution, sharing or resale. GitHub shows the license as Other. The EULA (updated 2026-04-17) grants a revocable, non-exclusive, non-transferable license and excludes safety-critical use.

Risks & limits

Keys and cloud sync. The pricing FAQ and the Privacy Policy say API keys are kept on the device and never sent to TypingMind; the cloud sync overview lists API keys and the license key among synced data.

Cloud sync encryption. The cloud sync docs describe HTTPS in transit and AWS encryption at rest, with no mention of end-to-end encryption; shared links are public, only kept out of search indexes; deleted items stay in Recently Deleted for 30 days; sessions expire after 30 days of inactivity.

Local data can be lost. In the default mode data lives only in the browser, so clearing browser data or switching devices loses history. The Privacy Policy also says cloud data of unpaid accounts inactive for 90 days is deleted after notice.

Vulnerability disclosure. The vendor's Vulnerability Disclosure Policy (docs dated 2026-10-02) sets the scope: typingmind.com, the cloud, eu, api and custom subdomains, and *.typingcloud.com; plugins and MCP servers added by users and customer self-hosted instances are out of scope. It promises acknowledgement within 3 business days, accepts anonymous reports and offers safe harbor.

Plugin permissions. Plugins can read chat attachments through the read_chat_attachments permission, while plugins and MCP servers sit outside the vendor's disclosure scope.

Provider dependence. The incidents page records several Anthropic API issues in March 2025, including failed cross-origin requests that were fixed by routing through a proxy; all are resolved. The Privacy Policy says the optional proxy passes requests through as-is without logging content.

Training and retention. The Data Usage policy says conversations are never used to train models, but provider rules still apply: the page lists 30-day default retention for OpenAI and Anthropic API data and explains how Gemini's paid and free tiers differ on training.

Contract terms. Disputes with non-consumer users go to arbitration at the Singapore International Arbitration Centre, with a class-action waiver; claims must be brought within 1 year; liability is capped at fees paid in the previous 12 months.

How it got here

OpenAI released the ChatGPT API on 2023-03-01; Tony Dinh registered the domain the next day and shipped the first version on 2023-03-06. The founder's newsletter (2023-04-02) says cumulative revenue passed $1,000, $2,000, $4,000 and $10,000 by 2023-03-10; on 2023-03-11 it was #1 Product of the Day on Product Hunt with 902 votes; revenue in the first 7 days was about $22,700, with 99% of sales from Twitter. All as reported by the vendor.

Pricing started with a $19 early-bird offer and rose to $39. The February 2024 newsletter reports $500K in total revenue, with the subscription side (cloud sync, Teams, knowledge base limits) at $15K MRR; the founder sees the mix of one-time and subscription revenue as a way to cover support for lifetime users. As reported by the vendor.

At 2024-02-09 22:54 (UTC+8) the GitHub repository TypingMind/typingmind was created, and the "Init" commit 3 minutes later brought in the static self-host package and LICENSE.md; the vendor tweeted the GitHub release of the static self-host version at 2024-02-11 09:58 (UTC+8). As of 2026-10-25 19:45 (UTC+8) the GitHub API showed 989 stars, 363 forks and 68 commits; recent commits are mostly "Automated update", the latest on 2026-10-10.

The newsletter of 2024-11-22 reports $1M in revenue over the previous 12 months, 20 months after launch, with one-time sales about half of monthly revenue. The focus then moved to business customers: one 3,000-seat deal and several 1,000-seat deals, branded reseller partnerships, and more than $20K spent on SOC 2, HIPAA and GDPR compliance and security audits. As reported by the vendor.

Later updates centered on Teams and model coverage: the EU data center, SSO and Directory Sync, audit logs, and steady support for new models. As of 2026-10-25 the pricing page claims more than 20,641 customers, while the README badge on GitHub still reads "Users: 18,000+"; both as reported by the vendor.

Our verdict

What sets TypingMind apart: it is a commercial bring-your-own-key chat front end with multiple models, agents, MCP, knowledge bases and team management; the personal license is a one-time payment of $39, $79 or $99; chats are kept in the local browser by default.

Its limits: the code is closed and the self-host package is compiled only; with cloud sync on, API keys and the license key are among synced data, and the cloud sync docs do not mention end-to-end encryption; there is no native mobile app; no tier includes model costs; the contracting party is Typing Technology Co. Ltd. in the Terms of Service but Devuap LLC in the checkout description.

Why it stands out

TypingMind is a chat interface that started as a weekend fix for a frustrating official UI and became a business Tony Dinh runs alone and reports on in public. Tony Dinh charges for the front end and lets you keep your own keys, which is the whole thesis.

Under the surface it is a study in interaction detail: folders, prompt libraries, plugins, multi-model conversations, and a local-first storage model that never sees your messages.

It is a chat interface that runs on your own API keys with local-first storage, built and run by one person, Tony Dinh.

Sources

Official

  1. TypingMind — Advanced chat UI for AI modelstypingmind.com · Accessed Oct 25, 2026
  2. TypingMind — Advanced chat UI for AI modelstypingmind.com · Accessed Oct 25, 2026
  3. TypingMind Changelogtypingmind.com · Accessed Oct 25, 2026
  4. TypingMind — Internal AI chat for your teamcustom.typingmind.com · Accessed Oct 25, 2026
  5. TypingMind Teams - Pricingcustom.typingmind.com · Accessed Oct 25, 2026
  6. Plans & Pricing - TypingMind Docsdocs.typingmind.com · Accessed Oct 25, 2026
  7. Manage License & Devices - TypingMind Docsdocs.typingmind.com · Accessed Oct 25, 2026
  8. Personal vs Teams - TypingMind Docsdocs.typingmind.com · Accessed Oct 25, 2026

Updates & reading