Peek is an open-source, infinite-canvas DB GUI with P2P multiplayer
Peek is a newly open-sourced database GUI built with Tauri (React + Rust) that displays queries and results on an infinite canvas, supports fully P2P multiplayer with live cursors, and includes an optional AI agent with MCP support.
Score breakdown
Peek's MCP integration lets Claude Code directly manipulate a live database canvas — creating nodes, moving the camera, and analyzing results — while its fully serverless P2P architecture means no connection details or table data ever leave the local network through a third-party service.
- 01Built with Tauri: React for the UI, Rust for DB interactions, LSP, and multiplayer networking
- 02Queries and results are displayed on an infinite canvas, similar to Figma or Miro
- 03Clicking primary or foreign keys fetches related rows as new nodes, building a directed acyclic graph
Peek is a database GUI client that tehrash has been building for roughly a year and recently open-sourced after using it as a primary DB client. It is built with Tauri, using React for the user interface and Rust for all database interactions, LSP functionality, and multiplayer networking. The core differentiator is that all queries and results are rendered on an infinite canvas — similar to Miro or Figma — rather than in traditional tabbed or grid views. Clicking a primary key surfaces all rows that reference it, while clicking a foreign key fetches and displays those related rows as new result nodes, progressively building a directed acyclic graph on the canvas. Additional canvas features include chart nodes for numerical data, whiteboard-style drawing, text nodes, and side-by-side query result comparison.
The multiplayer system is fully peer-to-peer, powered by iroh.computer, and works with live cursors in the style of Figma or Miro.
The multiplayer system is fully peer-to-peer, powered by iroh.computer, and works with live cursors in the style of Figma or Miro. Guests never receive connection details; instead, they send commands to the host and receive synced canvas updates in return. There is no centralized Peek service involved — not even for address resolution — so connection information and table data are not exposed to any third party. An optional built-in AI agent supports both local inference through any self-hosted OpenAI-compatible server (such as ollama or llama.cpp) and MCP integration, enabling tools like Claude Code to create nodes, move the camera, and analyze results. All AI features are opt-in, and MCP is disabled by default.
Key facts
- 01Built with Tauri: React for the UI, Rust for DB interactions, LSP, and multiplayer networking
- 02Queries and results are displayed on an infinite canvas, similar to Figma or Miro
- 03Clicking primary or foreign keys fetches related rows as new nodes, building a directed acyclic graph
- 04Fully P2P multiplayer via iroh.computer with live cursors; guests never see connection details
- 05No centralized Peek service — not even for address resolution
- 06Optional AI agent supports local inference (ollama, llama.cpp, etc.) or MCP integration with tools like Claude Code
- 07MCP is disabled by default; all AI features are fully optional
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Jun 15, 2026 · 11:57 UTC. How this works →