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.

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!

Currently, when importing models from OpenRouter into TypingMind (“Add Custom Model” → “Import OpenRouter”), users are required to manually set the “Model Capabilities” switches (System Role, Streaming, Plugins, Image Input, etc). This manual process is extremely error-prone and leads to substantial usability issues: If a capability is inadvertently set incorrectly, models can produce constant errors or do not function as expected in chat. OpenRouter already exposes a comprehensive, machine-readable list of supported capabilities and model properties via its API endpoint: https://openrouter.ai/api/v1/models (see also https://openrouter.ai/openapi.json). There is NO reasonable way to expect end-users to programmatically query the OpenRouter API, parse JSON, and map each capability manually to each TypingMind toggle. Feature Request: Please implement an automatic detection system in TypingMind: When importing models from OpenRouter, TypingMind should fetch the actual model metadata from https://openrouter.ai/api/v1/models (and/or the OpenAPI spec). The “Model Capabilities” switches in the import dialog should then be autoconfigured based on each model’s actual feature set, as published by OpenRouter (e.g., Streaming, Plugins/Tools, Image Input, System Role support, etc). If capabilities are ambiguous or missing in OpenRouter’s response, TypingMind should display a clear warning, but still do its best to pre-select what is documented. Manual override should remain possible for advanced users, but the default should always reflect the machine-readable truth from OpenRouter to minimize errors. Why is this important? TypingMind is otherwise not reliably usable with OpenRouter, as users must rely on guesswork or technical API analysis for every model import. This is a highly non-user-friendly experience and stands in contrast to TypingMind’s otherwise streamlined workflow. References: OpenRouter Model API: https://openrouter.ai/api/v1/models TypingMind “Add Custom Model” → “Import OpenRouter” dialog TypingMind “Model Capabilities” section Official OpenRouter documentation: https://openrouter.ai/docs TypingMind docs: https://docs.typingmind.com/chat-models-settings/use-openrouter-models Thank you for considering this essential usability improvement! The below screenshot demonstrates the core of the problem: When importing "Google: Gemini 3.1 Pro Preview", one of the most advanced models available, which is well-known to support nearly all modern features (system role, streaming, plugins/tools, image input, and more), almost all Model Capabilities toggles are disabled by default in TypingMind. Only "System Role" and "Streaming Output" are pre-selected. As a result, even though the underlying model can handle images, plugins, advanced reasoning, etc. these crucial capabilities remain unused and inaccessible unless the user manually enables each respective toggle. Manually enabling all toggles is not viable. Each model supports a different set of capabilities, and enabling unsupported features leads to constant errors. With multiple (or new) OpenRouter models, users cannot know all capabilities by heart and would need to check each model programmatically first. This is unmanageable and undermines the entire point of an integrated model import.

Hi TypingMind Team, I would love to use OpenCode's hosted models through TypingMind, but I'm running into a CORS issue. OpenCode's API endpoints do not send Access-Control-Allow-Origin headers, which prevents direct browser connections. I tried enabling the "Use TypingMind Proxy" option, but I received this message: "Sorry, TypingMind Proxy does not support the endpoint 'https://opencode.ai/zen/go/v1/chat/completions' yet." Could you please add the following OpenCode endpoints to the TypingMind Proxy whitelist? OpenCode Zen endpoints: https://opencode.ai/zen/v1/chat/completions https://opencode.ai/zen/v1/responses https://opencode.ai/zen/v1/messages https://opencode.ai/zen/v1/models/gemini-3.1-pro https://opencode.ai/zen/v1/models/gemini-3-flash https://opencode.ai/zen/v1/models OpenCode GO endpoints: https://opencode.ai/zen/go/v1/chat/completions https://opencode.ai/zen/go/v1/messages https://opencode.ai/zen/go/v1/models OpenCode is a very popular open-source AI coding agent (140K+ GitHub stars). Their Zen gateway provides curated proprietary models (GPT, Claude, Gemini, etc.), while GO provides affordable open-source coding models (DeepSeek, Kimi, Qwen, GLM, etc.). Supporting these endpoints would be a great addition for many TypingMind users who want reliable, tested models through OpenCode's gateway. If you need any additional details about their API format, you can find documentation here: https://opencode.ai/docs/zen https://opencode.ai/docs/go Thanks a lot for considering this!

Please have password protected conversation for shared computers

Hi, I think it would be great to have more options with TTS. And if it is possible please add a setting to manually input a OpenAI API endpoints, it would be great for user like me who host TTS models locally using AllTalk TTS and the likes. Thank you!

Your description on how to set up open router needs polished. When I enter my open router key I want to only see options that are usable on my chat dropdown menu when im trying other models. Many times I select a model and it says I can’t use it. I am not knowledgeable about these types of things so it could just be something im doing wrong but if not fix it. I enjoy the ease of access this app should grant me but i see clear paths for improvements.

Hey! I absolutely love the product and I've been using it for a few months a feature that I would really really love that openai Gemini and Claude have on their websites is saving memory automatically between chats or simply having every chat in an AI agent have some reference of what's going on in the other chat so every time I create a new chat I don't have to give the context or put it in system instructions. The MCP way to do this feels extremely tedious having to set up your own server I would suggest having an option to simply summarize what's going on in a chat and have every new chat have access to that. In terms of how to implement it that is obviously up to you guys I would suggest Maybe every time a new chat is created summarize two or three of the recent chats like the last 10 messages in the other recent chats and plug that in as part of the prompt. Thank you so much and this would be so appreciated!

Hi TypingMind Team, My name is Robert, and I am an AI Engineer and CTO at a biotech startup. I have been using TypingMind and really appreciate the solid tool you’ve built! I am writing to share a comprehensive feature proposal that I believe could elevate your platform to the next level, transforming it into a complete, multimodal AI hub. While I love the current integrations with OpenAI and ElevenLabs, I highly recommend expanding the platform's capabilities by integrating Google's latest preview models. Adding this suite of features would be a massive differentiator for power users, developers, and researchers. Here is my proposed suite of additions: 1. Google's TTS and STT Models Integrating Google's new audio models as options for both Text-to-Speech and Speech-to-Text would provide users with more versatility alongside existing APIs. I suggest looking into: gemini-2.5-pro-preview-tts gemini-3.1-flash-tts-preview 2. Video and Music Generation Models To make the platform a unified creative powerhouse, incorporating Google's media generation capabilities would be incredible. Specifically: Video Generation: veo-3.1-fast-generate-preview and veo-3.1-generate-preview Music Generation: lyria-3-clip-preview and lyria-3-pro-preview 3. Real-Time Conversational Mode (Voice/Video) I highly recommend adding a real-time, low-latency conversational mode (similar to the live modes in the ChatGPT and native Gemini apps). Using Google's gemini-3.1-flash-live-preview model would vastly improve the user experience for brainstorming, accessibility, and natural, fluid interactions. 4. Computer Use for UI Interactions Integrating agentic capabilities like Google's Computer Use (gemini-2.5-computer-use-preview-10-2025) would allow TypingMind to assist users with browser automation (this could be made via a Chrome Extension, for example), multi-step software tasks, and precise UI control directly within their workflows. 5. Deep Research Capabilities Finally, adding the Deep Research model (deep-research-pro-preview-12-2025) as a tool or plugin to Gemini Models would be a game-changer for academic and professional users. It would allow the platform to conduct autonomous, multi-step investigations, synthesizing complex information into comprehensive, properly cited reports based on web sources and workspace data. Thank you for your time and for continuously improving the platform! I’d love to hear your thoughts on these potential additions. I look forward to seeing how TypingMind evolves. Best regards, Robert Sousa Santos

I had a long image generation session and had to stop for bug reports. When I went back to chose the images I wanted to keep, the first 14 images had disappeared. I know you warn it is only for an hour, but can’t you keep them for the duration of the chat - or warn they are about to be deleted? I’d be ok if they was an option to always download every image I generate … better than losing work.

Request: Add the following endpoint to the TypingMind Proxy allowlist https://services.wolfram.com/api/agent-one/v1/chat/completions for the Wolfram Agent One API. This API enables computational, mathematical, and scientific grounding tools for LLM-generated responses to user queries. The API is fully compatible with the OpenAI Chat Completions API specification, including support for the standard messages format with system, user, and assistant roles, multi-turn conversations, and the same JSON response structure as OpenAI. I have successfully configured it as a Custom Model in TypingMind and verified that it works correctly when routing through a local CORS proxy. The only blocker is that the TypingMind Proxy returns the message: "Sorry, TypingMind Proxy does not support the endpoint yet." Wolfram Agent One documentation is available here: https://www.wolfram.com/apis/documentation/cag/wolfram-agent-one-api/ Thank you for your time and for building a great product.

When the local data gets corrupted, being able to delete the most recent chat session only, and from outside of TypingMind (say a CMD with a parameter following the executable), to hopefully save the remaining local data.

Google recently released the Gemini Deep Research Agent API, and I would like to request the integration of these models into TypingMind. These models utilize the same Google AI Studio API key as the existing Gemini series. There are a few ways this could be implemented: as a tool, a plugin, or—most effectively—as standalone models. Given that the Deep Research agent has unique search and reasoning behaviors, adding them as separate models seems to be the most robust approach. Please consider adding support for the following model IDs found in Google AI Studio: deep-research-max-preview-04-2026 deep-research-preview-04-2026 Enabling these would allow users to leverage Google's deep research capabilities directly within the TypingMind interface using their existing API configurations. Looking forward to seeing this feature in a future update.

Currently, TypingMind always adds a project context block like the following to requests, even if the project instructions field is empty: (No instruction provided for this project) This useless garbage is served along with every request in a project, consuming tokens and possibly confusing the model in a way that user a) may not want to, and b) has no control over. Request: If project instructions are empty or unset, TypingMind should not add any context block or annotation to the system prompt. If project instructions are set, TypingMind should insert ONLY the user's supplied text, with NO extra tags, wrappers, or comment text. The user should have 100% direct control over what goes into the prompt in the "Project Instructions" section.

Reasoning tokens can be a bulk of the context that gets passed there and back. Having it in the context is sometimes desirable, but often it’s not. Would it be possible to add an option to remove the reasoning_content from previous content from subsequent messages?

It would be so helpful to have the ability to share a chat with another user and the ability to collaborate with another user in a chat. Currently we are solving this with copying and pasting the entire chat into Google Docs in order to share it.

When importing models from OpenRouter, I need to re-enter my OpenRouter API key every time I want to add a new model. However, OpenRouter only shows the API key once when it is generated, and if I haven't saved it somewhere, I can't recover it to use in the future. This makes the process difficult if I forget to store the key, since TypingMind does not remember it. I would like TypingMind to securely save my OpenRouter API key after I enter it for the first time, and automatically use (or pre-fill) the same key whenever I import more models from OpenRouter. This would prevent me from losing access to my models if I lose the OpenRouter key, and make the import process much smoother. Thanks for considering this improvement!

Allow admins to set a fallback model that automatically kicks in when the primary model fails or hits a rate limit (e.g., Model A fails → switch to Model B). Useful for teams that need uninterrupted AI access in production.

Currently we can store prompts as templates, and these can reference variables in them. I’m looking for the opposite; store variables and reference them when I type out my prompts. The workflow I’m looking for is: save a named text block once, e.g. rules, style, brand_voice, disclaimer reference it while typing any prompt have TypingMind expand it inline before sending Example: ``` I need you to research frogs. Follow these rules: {{rules}} Using that research, create a Word document with this style guide: {{style}} Using that research, create a PowerPoint with this style guide: {{style}} ``` Why this would help: avoids repeating large blocks of text keeps prompts readable lets one shared block be updated once and reused everywhere reduces the need to maintain many near-duplicate prompt templates Current workarounds like full templates, agents, or system instructions are helpful, but they don’t solve the “define once, reference anywhere inline” workflow. Possible UX: a Snippet Library / Global Variables section invoke with {{name}}, /snippet, or a keyboard shortcut optional preview before send warning if a referenced snippet doesn’t exist This would complement Prompt Templates rather than replace them.

Hi Gerhard, Thanks for reaching out and for the clear explanation! Currently, the TypingMind Proxy only supports a limited number of official or widely-used API endpoints (e.g., OpenAI, Anthropic, OpenRouter, etc.). Unfortunately, third-party providers like aiprime.store are not yet supported through the TypingMind Proxy, which is why you see the “TypingMind Proxy does not support the endpoint https://aiprime.store/v1/messages yet.” error. What you can do right now: You can attempt to connect directly to the aiprime.store API from TypingMind without enabling the TypingMind Proxy. (Sometimes, this works, but due to browser CORS restrictions, this is often only possible in self-hosted/PWA or with desktop app versions.) If CORS or security issues block the direct connection, then unfortunately, there is no workaround within the current TypingMind Proxy system as it does not whitelist custom endpoints like aiprime.store. Feature request: We understand this could be a useful addition for users of alternate Claude API providers! I recommend Submitting a feature request here so our team can consider adding support for additional proxy endpoints in future updates. I'll also forward this feedback to our dev team. If you need more direct help, feel free to reply here or contact support@typingmind.com. Let me know if you have any other TypingMind questions!