mcpField
S
Stash: Remember Everything

Stash: Remember Everything

Remote
stash.graftedsoftware.comProductivity0Updated 3w ago

About Stash: Remember Everything

Stash MCP lets Claude search, read, and add notes to Stash, an iPhone notebook for organizing links, images, and thoughts. It can also answer questions across notes, a feature that requires Stash Pro.

Key features

  1. 01Search and open notes
  2. 02Create summarized notes
  3. 03Ask across notes with Stash Pro
  4. 04Update, delete, and research notes

Use cases

  1. 01Find saved details about a trip
  2. 02Save a to-do as a structured note
  3. 03Ask about flight information across notes

MCP support

A remote MCP server provides Claude with tools to search, read, create, and manage notes. Claude calls the tools using OAuth; question answering draws only on notes and does not use web search.

Configuration

Remote servers need no local install. Pick your client, then copy the config or install in one click.

Endpointhttps://stash.graftedsoftware.com/mcp
Copy install command
Claude CodeTerminal
Cursor.cursor/mcp.json
VS Code.vscode/mcp.json
Claude Desktopclaude_desktop_config.json
Windsurf~/.codeium/windsurf/mcp_config.json

Run commands in a terminal; paste JSON into the client’s MCP config and restart. Servers that need sign-in will prompt on first connect.

Tools

Details

From the project README

Stash

Stash is an AI notebook for iPhone. Share anything into it, such as a link, a screenshot, a photo, or a rough thought, and the Stash server turns it into a structured note with a title, summary, sections, tags, and sources. This plugin lets Claude read and add to that notebook.

What you can do

  • Search your notes. "Find what I saved about the Hawaii trip."
  • Open a note. Claude reads the summarized note before quoting it.
  • Save a note. "Stash this: renew the car registration next month." The server summarizes it once, in the background.
  • Ask across your notes (Stash Pro). "What did I save about flights?" Answers come only from your notes, with no web search.
SkillTools it calls
searchstash_search, stash_get
summarizestash_create, stash_get
askstash_ask

The server also exposes stash_me, stash_update, stash_delete, and stash_research. The skills only update or delete a note when you name that note and confirm.

Setup

  1. Get Stash from the App Store and sign in with Apple or Google.
  2. Install this plugin. The first time Claude calls a Stash tool, it opens a Stash sign-in page. Use the same Apple or Google account as the app.
  3. On the consent screen, review the scopes and select Allow.

Free accounts can search, read, and save notes. A free account gets 30 summaries and 5 research reports per month, resetting on the 1st (UTC). Ask needs Stash Pro.

What this plugin runs and sends

This plugin has no local code. It declares one remote MCP server, https://stash.graftedsoftware.com/mcp, run by Grafted Software, plus three skills written in Markdown.

  • Claude sends tool calls (search queries, note text you ask it to save, note ids) to that server, authenticated with an OAuth token you grant on the consent screen.
  • Notes are encrypted at rest (AES-256-GCM). They are not zero-knowledge: the server decrypts them to run tools, the same way it already does to summarize a note.
  • To summarize, research, and answer questions, the server sends note text to Anthropic (Claude), via OpenRouter. Summaries and research also use OpenRouter's web search. To decide whether a summary should use web search, it sends up to 2,000 characters of the note to TypeSafe (Jev), via OpenRouter.
  • Audit logs record the action, note id, and client id. They never record note bodies or search text.

To stop access, remove the connector in Claude or delete your Stash account in the app.

Privacy and support

License

MIT. See LICENSE.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.