Developer Docs
Connect your AI assistant directly to your knowledge.
The Pastepile MCP server lets Claude Code, Claude Desktop, Cursor, Windsurf, VS Code, and any Model Context Protocol client save, search, and reload knowledge on Pastepile. Durable memory for assistants, one command to install.
Connect in under 60 seconds
claude mcp add pastepile -- npx -y @pastepile/mcp
Then ask: "Save the text "hello from my assistant" to Pastepile and give me the URL."
Durable memory
Assistants save knowledge as URL-addressable pastes and reload it in any later session, on any machine.
Private by default
Unlisted saves, per-key isolation, and a server-enforced rule that encrypted content can never reach a model.
Zero dependencies
Pure Node 18+, no build step, auditable in one sitting. Nothing added to your supply chain.
Works everywhere
JSON-RPC 2.0 over stdio: Claude Code, Claude Desktop, Cursor, Windsurf, VS Code, Codex CLI, and more.
Getting started
Quick startConnect your AI assistant to Pastepile in under 60 seconds. Copy-paste setup for Claude Code, Claude Desktop, Cursor, Windsurf, VS Code, Codex CLI, Continue.dev, Cline, Roo Code, and any MCP client.AuthenticationThe Pastepile MCP server works anonymously, and an API key turns it into a private memory layer with per-key isolation and higher limits.ConfigurationEvery configuration surface of the Pastepile MCP server: two environment variables, team-shared config files, and self-hosted deployments.Remote endpointConnect over Streamable HTTP with no local process: the hosted MCP endpoint runs the same four tools as the stdio server, authenticated with your API key.
Core concepts
How Pastepile MCP worksArchitecture of the Pastepile MCP server: a stateless, zero-dependency stdio process that maps four tools onto the public REST API.Security modelWhat the Pastepile MCP server can and cannot touch: local execution, zero-knowledge encryption boundaries, and why protected pastes can never leak into a model context.Private memoryHow an API key gives your assistant a private, isolated memory namespace on Pastepile, and how scopes decide what search and list can see.
Tools
pastepile_saveSave text, code, or notes to Pastepile and get back a stable, shareable URL plus a one-time edit key. The write path of durable AI memory.pastepile_getRetrieve a paste by slug or full URL and reload previously saved knowledge into the model's context.pastepile_searchSearch pastes by keyword. With an API key it defaults to your private memory; without one it searches the public corpus.pastepile_listList recent pastes, newest first: your saved pastes with an API key, the public feed without one.
Reference
Error referenceEvery error the Pastepile MCP server can surface: model-visible tool errors, JSON-RPC protocol faults, and the underlying REST statuses.Rate limits and quotasExact request, size, and expiry limits for anonymous, free-key, Pro, and Enterprise usage of the Pastepile API and MCP server.Troubleshooting and FAQFixes for the failures people actually hit, plus straight answers about openness, offline use, data storage, and the remote endpoint.
Guides
Examples and workflowsReal prompts and the tool calls they produce: session memory, runbooks, agent handoffs, code review notes, and content pipelines.Ecosystem: CLI, editors, CIThe MCP server is one door into Pastepile. The CLI, VS Code extension, GitHub Action, and REST API read and write the same pastes.