New · MCP server

Your saved knowledge,
callable from any AI agent.

Stashwise MCP lets Claude Code, Codex, Cursor, and Claude Desktop search the library and wiki you’ve been building — with source-URL citations on every answer.

Set up in 60 seconds View on GitHub

$ claude mcp add -s user stashwise -- npx -y @stashwiseapp/mcp
$ npx -y @stashwiseapp/mcp auth

Three steps. No new tools to learn.

Save in Stashwise

Articles, videos, and social posts — saved from the web app at stashwise.co or the Stashwise iOS app while you scroll. Everything you save builds your library and wiki.

Add the MCP server

Drop one block into your agent's config, then run this once to pair your account:

$ npx -y @stashwiseapp/mcp auth

Ask any agent

Your agent now has a search_stashwise tool — grounded answers with citations to your sources.

What you can actually ask.

The MCP server returns ranked snippets from your library and wiki with the original source URL on every result.

Claude Code

“What did I save about HNSW indexes last month?”

Cursor

“Pull the API patterns I bookmarked into this project.”

Codex CLI

“Summarize my reading on Rust borrow-checker tricks.”

Set up in two commands.

Claude Code first — the most common host. Cursor, Claude Desktop, and Codex CLI work too (paste a snippet into their config file). Same package either way.

1. Install

run in your terminal

$ claude mcp add -s user stashwise -- npx -y @stashwiseapp/mcp

2. Authorize

run in your terminal

$ npx -y @stashwiseapp/mcp auth

A browser opens to confirm the pairing. Token stored in your OS keychain.

Using Cursor, Claude Desktop, or Codex CLI instead?

These hosts register MCP servers via config file rather than a CLI. Paste the snippet into the path shown, then run npx -y @stashwiseapp/mcp auth once from any terminal.

Cursor

~/.cursor/mcp.json

{
  "mcpServers": {
    "stashwise": {
      "command": "npx",
      "args": ["-y", "@stashwiseapp/mcp"]
    }
  }
}

Claude Desktop

~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "stashwise": {
      "command": "npx",
      "args": ["-y", "@stashwiseapp/mcp"]
    }
  }
}

macOS path shown. On Windows: %APPDATA%/Claude/claude_desktop_config.json.

Codex CLI

~/.codex/config.toml

[mcp_servers.stashwise]
command = "npx"
args = ["-y", "@stashwiseapp/mcp"]

Build a library that pays compounding dividends.

Every article, video, and post you save is one more thing every agent in your stack can recall.

Create your library Already a member? Set up MCP →

Your saved knowledge, callable from any AI agent.