Analysis updated 2026-05-18
Give an AI coding agent persistent clipboard history and named snippets it can reuse across sessions.
Cache expensive lookups so an agent doesn't repeat the same slow query.
Let an agent tag and annotate existing files without moving or copying them.
Build a small knowledge graph of facts an agent has learned, connected by typed relationships, for later recall.
| xfloukiex-lab/pandaclip | 0labs-in/vision-link | alfons-fhl/cursor-plan2api | |
|---|---|---|---|
| Stars | 4 | 4 | 4 |
| Language | TypeScript | TypeScript | TypeScript |
| Setup difficulty | moderate | moderate | moderate |
| Complexity | 3/5 | 3/5 | 3/5 |
| Audience | developer | developer | developer |
Figures from each repo's GitHub metadata at analysis time.
The optional desktop lens requires system Node.js 22.5 or newer for its read-only SQLite watcher.
PandaClip is a local tool that gives AI agents, like Claude, a place to keep track of things while they work: clipboard history, cached lookups, notes on files, and a small knowledge graph, all stored on your own computer with no cloud service and nothing running in the background unless a tool is actually called. It plugs into agent tools through the MCP protocol, a standard way for AI assistants to call external tools, and offers 40 tools grouped into four families. The clipboard family remembers copied text with expiration rules, lets you save permanent named snippets, and organizes them into labeled channels, while automatically refusing to store things that look like secret keys or tokens. The cache family is a simple expiring cache for expensive lookups an agent should not have to repeat. The bamboo family lets an agent add tags, notes, and metadata to files you already have without moving or copying anything. The garden family is a small knowledge graph where an agent can add facts as nodes, connect them with typed relationships, and search or trace through them later. Everything is stored as plain SQLite database files under a single home folder on your machine. Alongside the server, the project includes an optional desktop app built with Electron that shows a live, read-only feed of what your agents are doing, built by watching those same SQLite files for changes. It can also be pointed at your own notes folder, a folder of git repositories, or an existing knowledge graph, showing edits, commits, and new facts as they happen. The project deliberately does not include its own search engine, and its README suggests pairing it with a separate tool called magpie-search for that purpose, run alongside it as an independent MCP server. PandaClip is built by VektorGeist LLC and is released under the Apache License 2.0, with the PandaClip name and panda logo kept as reserved trademarks.
PandaClip is a local-first MCP server giving AI agents clipboard history, a cache, file tagging, and a knowledge graph, plus an optional desktop app showing a live feed of what agents are doing.
Mainly TypeScript. The stack also includes TypeScript, SQLite, MCP.
Apache License 2.0: use freely for any purpose, including commercial use, as long as you keep the copyright and license notices.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.