Openrouter supports an in-region routing setting, allowing users to specify that their data processing is limited to specific region (such as the EU). When an Openrouter user enables this, they will be allowed to use an endpoint such as https://eu.openrouter.ai/api/v1 for EU routing. However, the integration of OpenRouter in typingmind does not allow customizing the endpoint URL, meaning all calls will go to https://openrouter.ai/api/v1 instead. Currently I am only able to use in-region routing by using custom JSON imports or a shoddy vibe coded typingmind extension. It would be great if there were official support for endpoint URL modification.

I got the chatbot result and I tried to highlight some wordings to copy or reference to it. But it got frozen for my clicking features but not the scrolling. The entire window does not response to my clicking unless I refresh the website

It would be great to have a one-click button to summarize a long conversation and automatically start a new chat with that summary as the context—without having to copy and paste manually. This would help manage token limits, streamline the workflow, and keep conversations focused. Please consider adding this feature for a smoother chat experience!

I'd like to request that api.experientiallabs.ai be added to the TypingMind AI Gateway's supported endpoints. When I add it as a custom model and enable "Route requests via TypingMind Cloud", I get: "Sorry, TypingMind AI Gateway does not support the endpoint "https://api.experientiallabs.ai/v1/chat/completions" yet. Please contact support if you think we should add it."

Dear Sir or Madam, I am enthusiastically using the Skills feature in TypingMind. The only thing that bothers me, given the large number of skills I use regularly, is that they all have to be displayed individually and activated separately. Is it possible to add a function that allows skills, plugins, and MCPs to be grouped into separate folders, and perhaps also into shared function folders? A brief example: If I want to enhance scientific papers using AI, I need access to several databases via an MCP. I also use a plugin for creating diagrams and some writing style skills. To prevent these from being permanently enabled, I have to activate and deactivate them separately. In a bundled function folder, I could simply activate the entire folder with a single click. I hope I have been able to win you over to my proposal and thank you for your efforts in this regard.

Not sure if there a way to make voice persistent? Anyway to really use the agents or app in a more useful or interactive manner? Just keep hearing how this is a great system and I believe that’s true - I’m just not seeing the whole picture and I’m not sure how to bridge the gap? Unless I’m wrong and it just sucks which doesn’t appear accurate me- I believe it’s me missing something that is maybe even obvious?

Please add Capacities PKM connection

This is regarding the perplexity plugin. I am using Sonar to search for now, i see that there is a migration from Sonar to agent API where Sonar will be available thru 27 Sep. WIll typingmind's plugin be affected? Will it be updated from your side?

At present, it is possible to import available skills into TypingMind. It would be desirable if it were also possible to create your own skills in TypingMind. See the Langdock example.

Now we have skills, being able to invoke them with a forward slash like many other apps would be really helpful

Add IdP-Initiated Flow

When you write big prompts, it would be great so more of it. For example, if you write a big message in Slack the input box gets much bigger. In Typing Mind it’s just a little. It would be great to have more space to see your input prompt. The space can auto adjust based on your text of maybe even like big full screen view would be great.

Hi TypingMind Team, Regarding the Knowledge Base feature, I would like to request the ability to individually enable or disable specific knowledge bases. Currently, I might have around 10 knowledge bases covering different domains, but a specific task only requires 1 or 2 of them. Enabling all knowledge bases simultaneously introduces unnecessary noise that can degrade the LLM's response quality, and increases the risk of irrelevant/excessive context retrieval. Could you please add toggle switches (or a selector) to activate/deactivate specific knowledge bases independently? Thank you very much!🥰

Hello TypingMind Team, I would like to request the inclusion of Opper.ai as a native AI provider in the model settings, alongside OpenAI, Anthropic, OpenRouter, etc. Why this would be a great addition: - European-focused and GDPR-compliant: Opper.ai hosts and processes data in Europe, which is highly requested by enterprise and privacy-conscious users. - OpenAI-compatible API: Since their API already mimics OpenAI's structure, integration should be relatively straightforward. - Advanced orchestration: Opper offers great features for structured outputs and long-term memory that many of us are starting to use in our workflows. While it is currently possible to add it via the Custom Model option, having native support with automatic model fetching, proper iconography, and official token tracking would greatly improve the user experience. Thank you for considering this request!

Can we have it so that certain text can be highlighted, or at least have the font in a different color? Like “how about words enclosed within a block of quotation marks” turn a different color. Or Italicized text use another font color. Bold text another color etc etc etc.

It would be helpful to have an option to assign knowledge-base documents to a folder based on tags. This way, when working within a folder, the AI would not need to sift through all documents in the KB. We also would not be restricted by an AI agent assigned to a specific knowledge base. It would also be nice to have the AI update the documents in the knowledge base.

In fact, I have Claude’s new Cowork feature in mind.

Just a button to manually activate/make a context summary instead of having to always reach 70% of the context length.