Supermemory has a free plan that starts with $5 of credit, and Pro is $19 per month. The MCP server uses your Supermemory account, so usage counts against that plan. The repo is open source under MIT.

Does it support remote or OAuth?

Yes. It is remote only, at https://mcp.supermemory.ai/mcp, and clients discover the OAuth server through /.well-known/oauth-protected-resource/mcp. The server also accepts a Supermemory API key (sm_ prefix) as a bearer token.

How does it compare to a team memory service like Glen?

Supermemory stores what an agent or user saves into a space, and recall is a search over that space. Glen records every agent session and connected tool without a save step and makes the result searchable across the org, filtered per observation by what each user is cleared to see.

Supermemory MCP server

OfficialSupermemory31,588Config last verified Nov 10, 2026

Supermemory's hosted MCP server: save documents, search extracted memories, and read a profile per space, over OAuth at mcp.supermemory.ai.

Supermemory's MCP server is a remote endpoint at https://mcp.supermemory.ai/mcp, run as a Cloudflare Worker from the apps/mcp directory of the supermemoryai/supermemory repo. Content the agent sends through add_memory becomes a document; Supermemory then extracts memory entries from it asynchronously, and search_memory returns matching memories with a similarity score and the ID of the source document. get_profile returns stable and recent facts about the user for a space, and list_documents and get_document read the stored source material.

Memory is partitioned into spaces. Each call resolves its space from an explicit namespace argument, then the account's active space (kept in a Durable Object keyed by organization and user), then the account default. The server also ships MCP Apps: a space picker, a guided save form, a file upload form, and an interactive memory graph, which clients with MCP Apps support render inline. The repo is MIT-licensed. Writes happen when the agent or user calls a save tool, so what ends up in a space depends on the agent choosing to store it.

Quick install

Copy-paste configs are provided for all 8 supported clients. Pick your client below.

Add to ~/.claude.json

~/.claude.json

json

{
  "mcpServers": {
    "supermemory": {
      "type": "http",
      "url": "https://mcp.supermemory.ai/mcp"
    }
  }
}

Or via CLI

bash

claude mcp add --transport http supermemory https://mcp.supermemory.ai/mcp

Heads up

First tool call opens a browser to authorize.

Step-by-step guides

Available tools

Tool	Description
search_memory	Search memories in one space with a natural-language query; results carry a memory ID, similarity, and source document ID.
get_profile	Get stable and recent profile context about the user for a space.
list_documents	List document metadata, summaries, and processing status in a space.
get_document	Read one document's available content by ID.
list_memories	List extracted memory entries and the IDs of the documents they came from.
list_spaces	List the spaces visible to the authenticated account.
who_am_i	Return the caller's identity, access level, and active space.
add_memory	Accept content as a document for asynchronous memory extraction.
forget_memory	Forget a memory matching the supplied content.
select-space	Open the interactive space picker (MCP App).
memory-graph	Open the interactive memory graph (MCP App).
guided-save	Open the guided memory form (MCP App).
upload-file	Open the file upload form (MCP App).
set-active-tag	Persist the selected active space; called by the embedded app, hidden from the model.
save-memory	Submit the guided save form; called by the embedded app, hidden from the model.
prepare-file-upload	Prepare a direct file upload; called by the embedded app, hidden from the model.
fetch-graph-data	Fetch documents with their memories for the graph view; called by the embedded app, hidden from the model.

What you can do with it

Personal memory in ChatGPT and Claude

Connect the same Supermemory account to ChatGPT, Claude Desktop, and Cursor; get_profile and search_memory return the same saved context in each.

Separate spaces per project

Keep one space per client or codebase and pass namespace on a call to read another space without changing the active one.

Save files and notes from the chat

In clients with MCP Apps support, upload-file and guided-save open forms inline, so a PDF or a long note goes into a space without leaving the conversation.

FAQ

Is it free?: Supermemory has a free plan that starts with $5 of credit, and Pro is $19 per month. The MCP server uses your Supermemory account, so usage counts against that plan. The repo is open source under MIT.
Does it support remote or OAuth?: Yes. It is remote only, at https://mcp.supermemory.ai/mcp, and clients discover the OAuth server through /.well-known/oauth-protected-resource/mcp. The server also accepts a Supermemory API key (sm_ prefix) as a bearer token.
How does it compare to a team memory service like Glen?: Supermemory stores what an agent or user saves into a space, and recall is a search over that space. Glen records every agent session and connected tool without a save step and makes the result searchable across the org, filtered per observation by what each user is cleared to see.

Compare Supermemory alternatives →← Browse all agent-memory servers

MCP Server Directory Glen home