Claude Code + remote MCP
Claude can save it.
You can keep it.
Give Claude Code a small, explicit store for notes, todos and project references. Connect it to your account through MCP, then inspect what it actually saves.
Open-source core. Hosted Free plan. No card needed to start.
{
"key": "notes/release-checklist",
"kind": "note",
"labels": [
"release"
],
"value": {
"text": "Verify sign-in, billing and first save before launch."
},
"searchable_text": "Verify sign-in, billing and first save before launch."
}Illustrative object input—not a live account or tool result.
“Save this checklist to my notes.”
notes/release-checklist“Find my release checklist and show the saved text.”
store_object → search_objects → get_objectConnect the store, not a second chat app
Simple Agent Store exposes a Streamable HTTP MCP endpoint at https://www.simpleagentstore.com/api/mcp. Sign in, open Get started and create an access key for Claude Code. The dashboard gives you the client-specific configuration and connection test.
The key authenticates requests separately from your browser session. It gives the client access to your objects, so keep it in trusted local configuration rather than a prompt, shared screenshot or repository. Create separate keys for separate clients and revoke one when it is no longer needed.
Make a save you can verify
After connecting, check that Claude Code sees the five tools: store_object, get_object, list_objects, search_objects and delete_object. Approve tool use according to your client settings, then ask it to save a small test note.
Open the data viewer and check the key and original text. In a fresh conversation, ask Claude to retrieve that note through the store. A confident chat response is not proof that data was persisted; the object and the retrieval tool result are the evidence.
Use ordinary requests, with explicit storage intent
“Save this to my notes” or “add this to my todos” can guide a connected agent toward a write. The server provides routing instructions and the open-source project includes optional agent guidance. Those instructions help tool selection, but do not override permissions or guarantee a call.
This page focuses on Claude Code. Claude web and desktop connector support depends on the client’s available transports and authentication. Do not assume a connector supports this bearer-key endpoint simply because it accepts an HTTPS URL. Use the compatibility guidance in Get started for your actual client.
Three steps to a useful store.
- 01
Create a Claude Code key
Sign in and open Get started. Give the key a recognizable name and copy it securely.
- 02
Use the Claude Code setup
Select Claude Code in the workspace and use its remote HTTP configuration. Confirm the tools appear.
- 03
Save, inspect, retrieve
Save a small note, confirm it in the viewer, then retrieve it in a new conversation.
A few practical questions.
Will Claude automatically save every conversation?
No. Claude must call the storage tool. Only the explicit saved object is retained, not a conversation archive.
Can another agent retrieve Claude’s notes?
Yes, when it connects to the same account’s store and retrieves the saved key. It does not need Claude’s chat history.