# \[App\]\[Pro\] Mistral AI **URL:** \ **Category:** Apps\ **Tags:** homey-pro\ **Created:** [April 11, 2026, 4:31pm UTC](https://community.homey.app/t/app-pro-mistral-ai/153643 "2026-04-11T16:31:18Z")\ **Posts on this page:** 7\ **Page:** 1
**Author:** ![Thibaud\_Lgs](https://avatars.discourse-cdn.com/v4/letter/t/67e7ee/32.png) [@Thibaud\_Lgs](https://community.homey.app/u/Thibaud_Lgs)\ **Post date:** [April 11, 2026, 4:31pm UTC](https://community.homey.app/t/app-pro-mistral-ai/153643/1 "2026-04-11T16:31:18Z")
Hi there! I’ve been creating an app to use Mistral AI models directly on your flow cards > **[Mistral AI | Homey](https://homey.app/en-be/app/com.mistral.ai/Mistral-AI/test/)** > > Harness the power of Mistral AI LLMs directly in your Homey flows. # Setting up 1. install the app on your Homey pro or SHS 2. Go to mistral console (formerly “La Plateforme”) [https://console.mistral.ai/](https://console.mistral.ai/) 3. activate your “experiment plan” (free) or paid plan 4. create a new API key 5. paste the API key in the homey app setting 6. You’re all set ✌ # Features | feature | explanation | use case | recommended model | | --- | --- | --- | --- | | use Mistral LLM models | you can use the “ask mistral model” to generate text | create special notification or home summary | mistral 8b for small tasks and notifications, Mistral Medium is usually enough for more complex tasks | | use Mistral VLM models | you can use the “ask mistral model with images” to generate text | analyzing security camera snapshot | Mistral Small 4, Pixtral is not recommended | | use Mistral agent | you can create various AI agents on Mistral platform, you can select a model, temperatures, tools, customs instruction. You can even ask the model to use a specific JSON format! | | | | take action | The homey app will prompt mistral API with a list of your devices and their capabilities, to act in your smart home according to your initial prompt | create your own chatbot or with the new ask user homey notification | Mistral Large is strongly recommended Medium and Small 4 should work fine too | | conversation variable | At the end of each flowcard you can add a custom variable if you want to keep the conversation to add a follow-up request | chatbots or using another model for a secondary action | | # Disclaimer - many points of this app have been vibecoded and only then human supervised - If you use the free tier or “labs models” Mistral may use your request to train their models # Roadmap ## Ask Home Selecting whih devices are exposed to the AI and on which ones the AI has control Better context and commands Triggering flows User context ## Other TTS models Voxtral support ---
**Author:** ![cautje](https://sea1.discourse-cdn.com/flex025/user_avatar/community.homey.app/cautje/32/328_2.png) [@cautje](https://community.homey.app/u/cautje)\ **Post date:** [April 25, 2026, 8:31am UTC](https://community.homey.app/t/app-pro-mistral-ai/153643/2 "2026-04-25T08:31:29Z")
Great! It would be great to have a flowcard to upload a snapshot from a doorbel to mistral and get a discription back. Is that possible? Thx! ---
**Author:** ![Thibaud\_Lgs](https://avatars.discourse-cdn.com/v4/letter/t/67e7ee/32.png) [@Thibaud\_Lgs](https://community.homey.app/u/Thibaud_Lgs)\ **Post date:** [April 29, 2026, 11:22pm UTC](https://community.homey.app/t/app-pro-mistral-ai/153643/3 "2026-04-29T23:22:27Z")
Hi! This feature already exist btw, you can upload the image here as an attachment to your prompt. here is the flowcard in question. ![Capture d’écran 2026-04-30 à 01.16.00](https://us1.discourse-cdn.com/flex025/uploads/athom/original/3X/b/4/b4653561e1e388e31da998d1a47e32c0b8dd408b.png) But also it will be soon possible to attach an image to one of your custom agent, it will allow you to ask the models to respond with your specific instruction and/or with a specific JSON format. I could also create a spcific flowcard to analyze camera snapshot and return a description! Hope this helps! ---
**Author:** ![cautje](https://sea1.discourse-cdn.com/flex025/user_avatar/community.homey.app/cautje/32/328_2.png) [@cautje](https://community.homey.app/u/cautje)\ **Post date:** [May 1, 2026, 7:50am UTC](https://community.homey.app/t/app-pro-mistral-ai/153643/4 "2026-05-01T07:50:41Z")
Great! it works!!! THX!!! ---
**Author:** ![cautje](https://sea1.discourse-cdn.com/flex025/user_avatar/community.homey.app/cautje/32/328_2.png) [@cautje](https://community.homey.app/u/cautje)\ **Post date:** [May 1, 2026, 7:51am UTC](https://community.homey.app/t/app-pro-mistral-ai/153643/5 "2026-05-01T07:51:27Z")
I could use some examples, maybe we can share some? THX! ---
**Author:** ![cautje](https://sea1.discourse-cdn.com/flex025/user_avatar/community.homey.app/cautje/32/328_2.png) [@cautje](https://community.homey.app/u/cautje)\ **Post date:** [May 1, 2026, 11:41am UTC](https://community.homey.app/t/app-pro-mistral-ai/153643/6 "2026-05-01T11:41:45Z")
This works great! Thanks! I have another question: if you have multiple snapshots from different cameras described, how can you send the correct snapshot along with the answer? In the Mistral AI vision flowcard, you can’t select the original picture. So, the goal is to be able to read the Mistral response in, for example, a WhatsApp message and see the described image. That would be absolutely amazing. THX! ---
**Author:** ![4x4\_Pete](https://sea1.discourse-cdn.com/flex025/user_avatar/community.homey.app/4x4_pete/32/113286_2.png) [@4x4\_Pete](https://community.homey.app/u/4x4_Pete)\ **Post date:** [July 15, 2026, 7:49am UTC](https://community.homey.app/t/app-pro-mistral-ai/153643/7 "2026-07-15T07:49:42Z")
Hi @Thibaud_Lgs Is the source code publicly available? I am keen to learn, what MCP-API endpoints you have implemented. - Or does your App use the official Homey/Athom MCP Server? Background: I have a rather complex system prompt for Mistral based on the Homey/Athom MCP Server and tailored to my Homey system, that allows for complex status and event reports. It would therefore be interesting to see if that can be applied through your app as well and what has to be eventually changed. [EDIT] Works well. Found code on Github. [EDIT #2] Does not work. - Checked with test and live app version. Uninstalled, reinstalled, restarted (the usual dance). Mistral modell selection seems not to work. AI has no access to any device. See 4a023115-f058-4843-9e2a-f0e9202efa6d I have a Mistral pro account and the API key is valid and active. Would be cool if you could have a look.