Find answers in your documents
- BRING
- A project brief, report, or notes
- ASK
- Search your indexed documents with a question.
- GET
- A focused answer to review against the source.

Ask about your documents, turn data into charts, and plan everyday tasks.
AI Plate brings your chosen AI model and useful tools into one Windows app.
Windows x64 · Open source · MIT licensed · v1.0.1 · 293.2 MB
Choose a cloud provider with your own API key, or set up a local model.
Cloud usage may cost extra. Local models need separate downloads and suitable hardware.
Browsing on your phone? Explore the demo now and open this page on your Windows PC to install.
irm https://ai-plate.typobunch.com/install.ps1 | iexWindows x64 only. Downloads the desktop installer, verifies SHA-256, and opens the setup wizard.
macOS, Linux, and WSL are not supported by these scripts.

Universal Sovereign AI Agent Harness & Genius Butler
Balanced mode selected · Guided examples only. No prompts are sent to AI.
Interactive demonstration with sample data—not a live connection to your files or an AI model.
Pick one small task. Review the result. Build from there.
Available actions depend on your model, enabled tools, and permissions. You choose what to connect.
Install the app, choose a model, and try something small.
Download and open the desktop installer. Terminal commands are optional.
Use your cloud-provider API key, or configure a local server and model.
Ask a question about one document or create a chart. Review the result before expanding the task.
Explore skills for repeatable workflows, plugins for extra tools, and connectors for other apps.
Explore Marketplace Building your own? Read the developer guides →Understand the choices before you install.
Connect a supported cloud provider or run a model with Ollama or LM Studio. Chat and document embeddings are configured separately.
Find saved outputs in Artifacts and review personal memory in Settings. Relevant context may be sent to the provider you select.
AI Plate is MIT-licensed open-source software. Review the code and add capabilities through skills, plugins, and connectors.
A Windows x64 PC and a supported AI provider. Download the desktop installer, then configure a cloud provider with your own API key or set up a local model server. See the quick-start guide for configuration steps.
AI Plate is MIT-licensed open-source software. Cloud model providers may charge for requests. Running a local model requires suitable hardware and a separate model download. Marketplace listings may have their own terms.
Local chat and embeddings can run through local providers after models are downloaded and configured. Web search, external connectors, provider APIs, and downloads still need a network connection.
The app stores knowledge and personal memory locally. Content used in a request may be sent to your chosen chat or embedding provider. Choose local providers for both paths if you want local inference, and review enabled tools and connectors.
Tool access depends on the selected mode, enabled tools, and security settings. Start with a plan, review proposed changes, and inspect outputs. Python execution uses a workspace and subprocess; it is not a virtual machine that makes arbitrary code safe.
No. The preview uses sample data to illustrate workflows. It does not inspect your computer, send prompts to a model, or demonstrate measured performance. Install the desktop app to use real tools.
Open Docs → Desktop quick start. The guides cover providers, documents, voice, artifacts, memory, and extension development. Model availability and hardware requirements depend on the provider and model you choose.
Start with a document, a spreadsheet, or a question.
Download for WindowsWindows x64 · v1.0.1 · 293.2 MB · MIT licensed
Model setup required. Cloud-provider usage may cost extra.
Release details and available files ↗