simplenote-mcp
A remote MCP connector that gives Claude.ai projects a persistent, cross-conversation file store, backed by Simplenote and deployed free on Cloudflare Workers.
README
simplenote-mcp
A remote MCP connector that gives Claude.ai projects a persistent, cross-conversation file store, backed by Simplenote and deployed free on Cloudflare Workers.
The problem
A Claude.ai project conversation has no writable storage that survives across conversations — confirmed by probing the project sandbox from inside it:
| Surface | Agent can write? | Crosses conversations? |
|---|---|---|
/mnt/project |
no (only the user adds files) | yes |
/mnt/user-data/outputs |
yes | no — keyed per conversation |
VM (/home, /tmp) |
yes | no (ephemeral) |
/mnt/user-data/outputs is durable within a conversation, but its storage
namespace is the conversation id — a file written in one conversation is
invisible in the next, even inside the same project. The only cross-conversation
surface, /mnt/project, is read-only to the agent. So there is no
writable and cross-conversation surface — and because the sandbox's network
egress is allow-listed, a script running inside it can't reach Simplenote
directly either.
This connector supplies the missing surface from outside the sandbox: each
conversation pulls its files at the start and pushes changes back before ending.
The pull→work→push protocol the in-conversation agent follows lives in
BOOTSTRAP.md — add that file to your Claude project.
Why Cloudflare Workers
A connector must answer instantly when Claude calls a tool, which rules out
free hosts that sleep. Workers never sleep (~5 ms cold start), the free tier
needs no credit card, HTTPS is built in, and it's Anthropic's reference path
for remote MCP with OAuth (workers-oauth-provider + McpAgent). SQLite
Durable Objects — which McpAgent uses for session state — are now on the free
plan, so the whole thing runs for $0. (The sandbox's allow-listed egress can't
reach Simplenote, so the connector has to be remote anyway.)
The Simplenote/Simperium client is a faithful TypeScript port of the public
simplenote.py library (token auth,
the first-line-heading→filename convention, version-aware writes, markdown
system tag), so notes round-trip with the Simplenote apps.
Model
One Simplenote note per file, scoped by a project tag
(claude-project-<id>) that isolates a project's notes from the rest of the
account. A file's path is the slug of its first-line heading
(# Design Notes → Design-Notes.md).
| Tool | Purpose |
|---|---|
list_files |
discover persisted files |
read_file(path) |
pull one file |
write_file(path, content) |
persist/overwrite a file |
delete_file(path) |
trash a file |
Single-user by design: one Simplenote account (token stored as a Worker secret); OAuth just gates access to you with a shared password.
Deploy
Prereqs: a free Cloudflare account, pnpm + Node, and a
Simplenote API token. Mint one from your Simplenote email/password with the
public simplenote library:
pip install simplenote
python -c "from simplenote import Simplenote; print(Simplenote('EMAIL', 'PASSWORD').get_token())"
pnpm install
# 1. Log in to Cloudflare (opens a browser):
pnpm exec wrangler login
# 2. Create the KV namespace the OAuth provider uses, then paste the printed id
# into wrangler.jsonc (kv_namespaces[0].id).
pnpm exec wrangler kv namespace create OAUTH_KV
# 3. (optional) set your project tag in wrangler.jsonc → vars.SIMPLENOTE_PROJECT_TAG
# 4. Set secrets (not in any file):
pnpm exec wrangler secret put SIMPLENOTE_TOKEN # your Simperium token
pnpm exec wrangler secret put ACCESS_PASSWORD # a password you choose
# 5. Ship it:
pnpm run deploy
You'll get https://simplenote-mcp.<your-subdomain>.workers.dev.
Connect in Claude.ai (Max/Team/Enterprise): Settings → Connectors → Add
custom connector → URL https://simplenote-mcp.<subdomain>.workers.dev/mcp.
Claude runs the OAuth flow; enter your ACCESS_PASSWORD on the login screen.
Finally, upload BOOTSTRAP.md into the project's files
(open your project → Add files) so every conversation picks up the protocol.
Then, in a chat, ask Claude to load my project files from Simplenote to
rehydrate at the start, and save these back to Simplenote to persist before
you finish.
Local development
cp .dev.vars.example .dev.vars # fill in SIMPLENOTE_TOKEN + ACCESS_PASSWORD
pnpm dev # wrangler dev (local Worker)
pnpm dlx @modelcontextprotocol/inspector # point at http://localhost:8787/mcp
pnpm run typecheck # tsc --noEmit
pnpm test # vitest: store/simperium logic (in-memory, no network)
Layout
src/index.ts— OAuthProvider wiring (the entrypoint)src/mcp.ts—McpAgentDurable Object + the four toolssrc/store.ts— file↔note mapping, tag scoping (port ofstore.py)src/simperium.ts— Simperium HTTP client (port ofsimplenote.py)src/auth.ts— single-user OAuth login gateBOOTSTRAP.md— the doc you add to the Claude projectpython/— optional local stdio variant for Claude Desktop / Claude Code
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.