← All products

Platform

A local-first AI assistant platform built on the DeepSeek Harness runtime, with streaming agent chat, document RAG, and MCP extensibility across web and desktop.

壹Base· AI assistant platform

▶ Try it live

Platform (v1.3.0) is the execution base of the FindData product matrix — the Base line (寻数·壹座): every FindData project runs on this AI harness. Built on the DeepSeek Harness (dsh) runtime, it ships as a web app and as an Electron desktop client — macOS .dmg and Windows .exe — with the same UI on both ends, and a private-deployment edition is available.

Streaming agent chat

Conversations stream in real time and render the full agent loop: chain of thought, tool calls, and skill execution. For each conversation you can choose the model and the agent to run it with.

Document RAG on a local index

Bring your own PDF, Markdown, and plain-text files, or point Platform at web URLs. It builds a local index over them and retrieves from that index during the conversation, so answers are grounded in your documents.

Agents, MCP, and scheduled work

The platform surface around the chat covers:

Local-first by design

A single command starts the whole thing, and all data lives in local SQLite. Keys and tokens stay on the server side — they never enter the browser. The interface is Chinese-first, with en/es/fr/ja also available.

Try it

Open the demo.