Meet your AI workspace
Set up your butler, explore the desktop features, and extend AI Plate with your own tools.
What would you like to do?
AI Plate combines conversation, tool-assisted work, document retrieval, and local voice in one desktop app. Start with a working provider, then add the capabilities your task needs.
Configure a provider and complete a small first task.
Read guide →Choose a working modeUnderstand Agent, Plan, Chat Only, and thinking depth.
Read guide →Work with your documentsChoose between structured knowledge and semantic retrieval.
Read guide →Speak and listen locallySet up Moonshine speech input and Kokoro speech output.
Read guide →Find your deliverablesSort, preview, and open files, audio, and video.
Read guide →Manage personal memoryReview facts, reflections, and your local dossier.
Read guide →Current desktop capabilities
| Capability | What it does |
|---|---|
| OKF + vector RAG | Combines structured knowledge nodes and relationships with document chunks retrieved by similarity. |
| Force Graph | Explore connections and inspect nodes in the Knowledge Base. |
| Auto-Distill | Extract a draft OKF node from a file or pasted text using local heuristics. Review the result. |
| Code-link verification | Checks whether referenced paths exist. It does not prove code correctness or rule compliance. |
| Unified Agent mode | Answers directly or uses enabled tools, subject to mode and security settings. |
| Artifacts & media | Six sorting options, image/code previews, and audio/video playback. |
| Mistral and local endpoints | The registry includes Mistral, Ollama, and LM Studio; chat and embedding configuration are separate. |
Extend AI Plate
| Surface | Use it for | Where it runs |
|---|---|---|
| Skill | Reusable instructions, tone, and workflow constraints | Added to model context when active |
| Plugin | Tools, artifacts, storage, and UI extensions | Desktop runtime or renderer |
| Connector | External services or desktop companion bridges | Service outside the app |
| AI provider | Model responses; embeddings when supported | Selected local or cloud endpoint |
How a request travels
Message + session mode + applicable skills
→ relevant memory and retrieved knowledge
→ selected model
→ allowed, enabled tools (when needed)
→ security and approval checks
→ tool results → model response → saved outputsA mode controls which tools can be offered. Enabling a plugin does not bypass that mode or its approval settings. Chat, retrieval, and tool execution can use different services.
Scope and verification
These product guides were checked against the local desktop checkout reporting version 1.0.0 on October 5, 2026. Implementation sources are listed in each guide. This is a source review, not a claim that every provider or model was tested live.
Implementation sources
Paths in your AI Plate source checkout:
package.jsoncore/chat-mode.tscore/orchestrator.tsui/index.htmlui/app.js