Store your knowledge once. Every AI tool remembers.
One MCP endpoint. Full OAuth. Semantic search.
$ claude mcp add --transport http memory-mcp https://mcp.devfrend.com/api/mcp
Connecting... OAuth consent in browser...
Connected to Memory MCP (8 tools available)
$ "What's my tech stack?"
search_memory("tech stack")
Next.js 16, Tailwind v4, Supabase, Vercel — 94% match
✕ Repeat your tech stack every session
✕ Re-explain coding rules to every tool
✕ Lose project context between conversations
✕ Different AI tools, different starting points
✓ Store once, every tool knows
✓ Rules loaded automatically via get_rules
✓ Full project context in one call
✓ Same brain across Claude, Cursor, Codex
Works with every MCP client
Claude Code
claude mcp add --transport http memory-mcp https://your-domain.com/api/mcp
Claude Chat
Settings → Connectors → Add Custom Connector
Cursor
{ "mcpServers": { "memory-mcp": { "type": "http" } } }
Codex
codex mcp login memory-mcp
Add memories via dashboard or directly from AI tools. Rules, preferences, project context.
One command. OAuth handles auth. No API keys, no config files, no friction.
Your AI tools read memories automatically. Semantic search finds anything by meaning.
Everything you need. Nothing you don't.
Rules, preferences, decisions, snippets — stored once, used everywhere.
Find by meaning, not keywords. Powered by pgvector embeddings.
Visual .claude.md manager. Export coding standards as markdown.
Scope memories to projects. One tool call loads everything.
Claude, Cursor, Codex — one endpoint, one OAuth flow, all clients.
Your Supabase, your data, your rules. Open source, MIT licensed.
# connect to the hosted instance — no setup needed
$ claude mcp add --transport http memory-mcp https://mcp.devfrend.com/api/mcp
# authenticate in browser, then you're done
# or self-host it yourself
$ git clone https://github.com/theamargupta/memory-mcp.git
$ cd memory-mcp && npm install && npm run dev