Memory, native to your editor.
Connect Cursor and other AI editors straight to Memuron over a streamable HTTP MCP server — authenticated with Clerk OAuth. No keys to copy, no glue code to write.
What deploy preferences has this user set?
semantic "deploy preferences"
| grep -i "release"Found 2 memories: prefers concise launch notes, and dislikes emojis in changelogs.
The tool surface
Every action is exposed as a typed MCP tool with safety confirmations on destructive calls — the same surface the HTTP API offers, flattened for agent tool pickers.
Discover the tool surface and the graph-filesystem query manual.
Pipe semantic search, traversal, and regex over the memory graph.
Queue raw content for the Guardian to merge or create.
Fetch a single memory and its links by ID.
Patch content or scope on a known memory, guarded by confirm.
Remove a memory — requires an explicit confirmation flag.
Poll an async ingest job until it completes or fails.
List spaces, set a default, and enable or disable per user.
Resolve any chunk or image back to its original source file.
Connect your editor
From zero to recall in under a minute — add the server, sign in with Clerk, and let your agent read and write the memory graph over the stream, scoped to your org.
Add the server
Point your editor at the Memuron MCP endpoint — one line of config.
Sign in with Clerk
OAuth with PKCE and dynamic client registration. No API keys to paste.
Remember everything
Your agent reads and writes the memory graph over the stream, scoped to your org.
{
"mcpServers": {
"memuron": {
"url": "https://memuron-production.up.railway.app/mcp",
"transport": "http"
}
}
}