Model Context Protocol
Give Claude Code, Claude Desktop, Codex and Cursor 38 tools over your notes, AI models, scheduled tasks and agent connections. One command to install, no API key.
claude mcp add pandanpc -- npx -y @pandanpc/mcp-serverWorks on Linux, macOS (Intel & Apple Silicon) and Windows. The binary is fetched on first run.
Your MCP client speaks the protocol over stdio. The server translates those calls into your PandaNpc account — the same notes and settings you see on the web.
Grouped by what they touch. Every name below is what the model actually sees in tools/list.
Read, write, search and reorganise PandaNote documents. Markdown in, blocks out — headings, tables, todo lists with nesting, code blocks and Mermaid diagrams all round-trip.
Register providers, add or retire models, inspect what is configured — so the agent can manage its own model catalogue instead of you clicking through a dashboard.
Cron-style jobs that run an agent on a schedule. Create them, edit prompts, toggle them on and off, and read past runs — all from inside a conversation.
Sync instruction files across machines and record what was done. Useful when several agents share one repository and need the same guidance.
List and create the connections that let an agent reach your machines.
| Client | Setup |
|---|---|
| Claude Code | claude mcp add pandanpc -- npx -y @pandanpc/mcp-server |
| Claude Desktop | Add to claude_desktop_config.json under mcpServers |
| Codex | [mcp_servers.pandanpc] command = "pandanpc-mcp" in config.toml |
| Cursor / other MCP clients | Any client speaking MCP over stdio |
MCP is an open protocol that lets an AI assistant talk to external systems through a uniform interface. Instead of every tool inventing its own plugin format, a client such as Claude Code speaks MCP to any number of servers. Each server advertises a list of tools; the model decides which to call and the client executes them.
Run claude mcp add pandanpc -- npx -y @pandanpc/mcp-server. The npm package downloads a platform binary on first run (Linux, macOS Intel and Apple Silicon, Windows). Then run claude mcp list to confirm it is connected.
No API key. The first call asks you to sign in with an email verification code, and the credentials are cached locally for 90 days. Nothing is stored in your shell history or config files.
Notes are stored in your PandaNpc account, the same ones you see on the web and in the desktop and iOS apps. The MCP server is a client of that account, not a separate copy — an agent writing a note updates the document you already have open.
Yes. Note edits go through incremental sync rather than whole-document replacement, so two agents editing different parts of the same note do not clobber each other.
claude mcp remove pandanpc. Nothing is left behind except the cached credentials file, which you can delete at ~/.pandanpc-mcp/credentials.json.
claude mcp add pandanpc -- npx -y @pandanpc/mcp-server