Many of us have needs for bigger storage than 50MB, and we all have already paid subscriptions for like Google Drive or Dropbox, and thus for would be nice if this option would be available to choose from instead to be bound to your Cloud. Thansk!

Hi Tony, Thank you for the update on expanded model provider support and response API endpoint support. I am very happy to see my long-standing request has been fulfilled. Please have option to display Credit balance under api key management, many provider have endpoint to get credit balance details Like for vercel: Base URL: https://ai-gateway.vercel.sh/v1 Endpoint: GET /credits Json Key: balance. for openrouter: curl https://openrouter.ai/api/v1/credits -H "Authorization: Bearer " { "data": { "total_credits": 100.5, "total_usage": 25.75 } } Thank you for the continued improvements to TypingMind.

I would like TypingMind to support connecting to Ollama Cloud using its official OpenAI-compatible API endpoint (https://ollama.com/v1/). Ollama Cloud offers hosted versions of models like Llama, Gemma, Nemotron, but currently TypingMind’s proxy does not support this endpoint, and browser CORS blocks direct access. Please add proxy and/or native support for Ollama Cloud’s API endpoint, so that users can use Ollama Cloud models within TypingMind just like OpenAI or Mistral. This would enable access to the latest open-source models with easy paid hosting. Thank you!

I hope there is a button to allow Multi-agent Debate Mode. This is I think even more useful than parallel agent output. MAD can save so much time analyzing a complicated topic, especially those with many sub-topics (ex: should universal basic income be implemented in the next years?) [MECHANICS] - Two or more AI language models debate a user-provided topic through structured, round-based interaction. - The human user: a) observes all outputs at every stage; b) may intervene between rounds; c) remains the final arbiter of whether consensus has been reached. [FEATURE REQUEST] - Please add a file attachment tool that allows the human user to upload and provide multiple reference files simultaneously to all participating LLMs. - The file attachment tool should support, at minimum: 1. PDF files 2. .txt files 3. Markdown files 4. docx/MS Office/LibreOffice files - The attached files should be available as shared reference material for any/all LLMs during the debate. - The human user should be able to attach files before the debate begins and, ideally, also between rounds when asking follow-up questions or requesting clarification. ** A strong multi-LLM debate protocol should not imitate a loose human panel discussion. It should behave more like a formal academic debate with controlled rounds, visible transcripts, forced engagement. - A multi-agent debate protocol should also offer the option to: - have the MAD protocol decide whether there has been full consensus on the entire list of topics, or - allow a human judge to decides whether true consensus has actually been reached on any or all of the discussion points.

I didn't expect them to restrict me to using only their cloud storage, give me only 50MB, and force me to pay a monthly fee. Is there any way you can sell me 1GB of lifetime storage or free up space for my existing cloud account? I have several cloud services; it seems absurd to have to pay a monthly fee for lifetime storage.

Hello TypingMind Team, I would like to request support for adding Atlas Cloud as a full custom provider in TypingMind. At the moment, TypingMind only allows adding individual custom models manually. This works, but it becomes inconvenient when using providers like Atlas Cloud that expose multiple models through an OpenAI-compatible API. I would like to be able to add Atlas Cloud once as a custom provider, using = Base URL: https://api.atlascloud.ai/v1 Then TypingMind should automatically fetch and display all available models from that provider, instead of requiring each model to be added manually one by one. This would be very useful for users who use model aggregators or OpenAI-compatible providers, because models are frequently added, renamed, or updated. Having automatic model discovery would make the setup much faster and easier. Suggested feature: Add a new option called Add Custom Provider Allow users to enter a provider name, API key, and base URL Automatically fetch available models from the provider’s /models endpoint Let users enable or disable the models they want to use Keep the current manual “Add Custom Model” option for advanced cases This would make TypingMind much more flexible for users who rely on external providers such as Atlas Cloud, OpenRouter-style services, or other OpenAI-compatible APIs. Thank you for considering this feature. Best regards, Fabien

Currently, TypingMind does not support training models directly based on your chat history or previous conversations. Please add this feature. It is very important

In short: I experience severe performance issues/lags when chats exceed a certain length, making the app painfully slow or impossible to use at that point. If this problem is shared by others and due to limitations of a PWA-style app, I would strongly recommend moving towards a native desktop (macOS) app. Problem: Unlike a similar post, I do not experience the problem as one of the number of chats rather than of chat length: As soon as a chat’s context exceeds ~20.000-30.000 Tokens and this particular chat is opened, every new action (clicking on this chat, scrolling within the chat, formulating a new question, querying an answer) will cause a ~10-20 second delay/freeze of the app. Unlike that other post, too, this is not accompanied by a huge CPU or memory usage (percentage-wise) of the app. Context: Your support bot attributes this to limitations of both the browser / the PWA using the browser’s local storage (IndexedDB), which I think is convincing. However, because I see no other complaints like this in this forum, I will share some additional context information to rule out other causes: Account: Typing Mind Premium Plan; Web application, locally stored data (no cloud). App: Typing Mind PWA operating on Safari 26.5 System: Mac OS Sequoia on a 3,2 GHz 6-Core Intel Core i7 (outdated, but usually not a problem; see also low CPU and memory usage) Internet: 300Mbit/s Download and 150Mbit/s Upload Closing comment: I really enjoy Typing Mind’s idea, UI, and possibilities. It is simply that bumper that makes the application so hard to use in the long term. I would be delighted to have a smooth performance without constantly splitting up chats (losing their context, too) or switching to the cloud – be it through a native desktop application or another workaround. Thank you!

Currently we can add MCP via json under MCP settings which will be severed under plugin labelled as MCP. Here can able to use local or remote server not typingmind cloud server. And it's legacy to show under plugin. Have better MCP management with all different transport type. 1.stdio 2.stdio + http 3.http or sse

Add private/API access to update AI agents and prompts programmatically. This would help teams manage multiple agents more efficiently and automate their workflows.

I am reporting a connection issue with the Kilo AI Gateway. It currently lacks CORS support, which prevents it from working directly within the TypingMind browser interface. Can support for Kilo Gateway be added to the TypingMind proxy? This would greatly improve the workflow for users relying on Kilo’s models. Thank you for thinking about this! Best regards,

Can you make it possible to record for longer than 2 minutes for the speech to text transcription and to minimise the recording window , as currently it obscures the conversation. This would enable me to read through the reply as i respond. Id also appreciate actually being able to listen to longer responses - there appears to be quite a short limit on both which renders the text to speech feature useless to me as my input is always longer than the limits. This has resulted in me having to rely on external apps such as Spokenly and Speech central in order to gain the functionality that i was used to on the native GPT app. I would appreciate not having to juggle so many apps and fragment my experience just to achieve that.

As of right now, the only viable option is to have a custom plugin with the current 1 hour self image deletion and this a TypingMind feature so would TypingMind consider up to 1 day for manual saving or a native way to automatically save generated/uploaded images to TypingMind’s cloud servers that doesn’t require a custom plugin. As a creative who frequently uses image generations this method would prevent an unnecessary rush to save all images within a set time frame whilst trying to multi-task including writing/talking to an AI, save images and interrupting one’s creative workflow. In saying that, would TypingMind kindly improve and expand the current image generation self delete time especially for creatives or at least allow for self-saving onto your TypingMind server. Also, will be an opportunity for Anthropic, Claude users that the prompt caching TTL can be expanded to 1 hour natively via a slider of some sorts as the current ~5 minutes (depending on API configuration) creates disruption in longer creative or design workflows where context reuse across extended sessions is required or that you use voice to text or actively talking to an AI that you need more than 5 minutes to send your message. In saying that, kindly consider making a slider option for users who prefer ~5 mins or 1-hour prompt caching TTL rather than an optional or inconsistent configuration dependent on backend routing or API limitations. Thank you.

I want to give one suggestion that in chats there is a search chat option when we open this,we see column in that we see questions that we have asked in the chat before but when we click on these questions, we are not going directly to that question in the chat. please put this feature in it is very helpfull.

Please bring support to newer speech models for voice input.

This would be so helpful as it will allow us to have model diversity - we choose our own model based on the nature of content we are browsing. Alternatively, if TypingMind is planning its own browser, that’s also great. Kindly have a thought on this.

Currently in TypingMind there are options to send All Messages for context or 1 to 100 messages. Since I use agents as relational and therapeutic purposes from time to time, the 100 message limit is too short for that purpose, but All Messages gets heavy fast and hits capacity too quickly. I'd like a way to set the limiter to 200 or 300 so the rich context is still there but doesn't hit capacity.

When using multi-model dialogue in TypingMind, each model's response varies in length, causing the conversation columns to fall out of sync after a few rounds. The suggestion is to align columns by the user's input message (rather than the model's response), so all models stay visually in step. Ideally, this would be a toggleable option in settings to suit different users' preferences.

Hi there, I would like to request the ability to customize the cache timeout setting for Claude models within TypingMind. Currently, the cache timeout seems to be hardcoded to the default 5 minutes. In my workflow, this duration is too short, resulting in frequent cache misses and inefficient usage of the feature. Since the Claude API supports a cache timeout of up to 1 hour, it would be extremely beneficial if we could toggle or configure this setting within the Model settings (e.g., a selection between 5 minutes and 1 hour). Having the flexibility to extend this timeout would significantly improve performance and cost-efficiency for long-running sessions or complex tasks. Could you please consider adding this configuration option? Thank you for your hard work on this project.