membase

What is Membase?

AI agents are stateless by default. Every new session starts from zero — no memory of who you are, what you've worked on, or what decisions you've made. You re-explain yourself constantly. Context gets lost. Agents get dumber the more you use them across different tools.

Membase fixes this. It's a universal memory layer that sits between you and every AI agent you use. It automatically captures context from your conversations and connected apps, structures it into a living Knowledge Graph, and delivers the right context to any agent the moment it needs it — without you lifting a finger.

Your agents stop forgetting. They start knowing you.

The Problem

How It Works

Connect → Capture → Route → Retrieve

1. Connect — Add Membase to any agent with a single MCP command. Works with Claude, ChatGPT, Cursor, VS Code, Codex, and more.

2. Capture — As you work, Membase automatically processes your conversations and pulls data from connected apps (Notion, Slack, Gmail, GitHub, Google Drive, Google Calendar) into your personal memory store.

3. Route — Incoming context is intelligently routed to the right store:

• Memory — Personal context, preferences, and history organized as a Knowledge Graph
• Wiki — Factual reference material as linked markdown documents

4. Retrieve — When an agent needs context, it calls search_memory or search_wiki. Membase returns the most relevant information using hybrid retrieval — not just what's recent, but what's actually useful.

Features

🧠 Living Knowledge Graph

Your memory isn't a flat list of facts. Membase structures everything you know into a connected graph — nodes link to related nodes, so agents can traverse context the way you actually think. Every new conversation updates the graph automatically.

⚡ Context-Aware Agent Delivery

Membase delivers your context to every agent at runtime. Agents call search_memory and search_wiki to pull exactly what they need — reducing token bloat, improving response quality, and making every interaction feel personal.

🔄 Multi-Source Sync

Connect once, capture everywhere. Membase ingests from:

• Chat history — ChatGPT, Claude, Gemini
• Productivity — Notion, Slack, Google Drive, Gmail, Google Calendar
• Dev tools — GitHub
• Knowledge bases — Obsidian Vault import

🔌 MCP-Native

Built on the Model Context Protocol. One line to integrate. Works with any MCP-compatible agent — no custom plugins, no per-tool setup.

🗄️ Hybrid Memory Retrieval

Membase doesn't do naive vector search. It combines:

• Vector search — semantic similarity
• BM25 — keyword relevance
• Graph relationships — connected context traversal
• Recency weighting — recent events surface first
• RRF (Reciprocal Rank Fusion) — combines all signals into a single ranked result

This means agents get context that is semantically relevant, keyword-matched, structurally connected, and timely — all at once.

📋 Dual Memory Stores

Supported Agents & Platforms

Data Sources: Notion · Slack · Google Drive · Gmail · Google Calendar · GitHub · Obsidian

Quick Start

Add Membase to any MCP-compatible agent with a single command:

# Claude Desktop / Claude Code
# Add to your MCP settings:
{
  "mcpServers": {
    "membase": {
      "command": "npx",
      "args": ["-y", "@membase/mcp"]
    }
  }
}

Then connect your data sources and let Membase start building your Knowledge Graph automatically.

→ Full setup guide

Architecture

┌─────────────────────────────────────────────────────┐
│                    Your AI Agents                    │
│         (Claude, ChatGPT, Cursor, VS Code...)        │
└────────────────────────┬────────────────────────────┘
                         │ MCP
                         ▼
┌─────────────────────────────────────────────────────┐
│                     Membase                          │
│                                                      │
│  ┌──────────────────┐    ┌────────────────────────┐ │
│  │   Memory Store   │    │      Wiki Store        │ │
│  │  Knowledge Graph │    │   Linked Markdown      │ │
│  │  search_memory() │    │    search_wiki()       │ │
│  └──────────────────┘    └────────────────────────┘ │
│                                                      │
│  Hybrid Retrieval: Vector + BM25 + Graph + RRF       │
└──────────────┬──────────────────────────────────────┘
               │ Auto-capture
               ▼
┌─────────────────────────────────────────────────────┐
│                   Your Data Sources                  │
│  Notion · Slack · Gmail · GitHub · Drive · Calendar  │
└─────────────────────────────────────────────────────┘

Pricing

→ View pricing

What People Are Saying

"This is the one memory layer that actually works across all my agents."

— Ignacio Montenegro, Entrepreneur

"It feels like a real Jarvis for my work, remembering context across projects."

— Minsung Park, CEO

Recognition

🥈 Claude Code Hackathon — 2nd Place (November 3, 2025)

The Universal Memory MCP that became the foundation of Membase took 2nd place at Anthropic's Claude Code Hackathon. It's been evolving into what Membase is today ever since.

Use Cases

Founders — Keep agents updated on your company context, investor relationships, and strategic decisions without re-explaining every session.

Developers — Give coding agents persistent knowledge of your codebase architecture, conventions, and past decisions.

Researchers — Build a second brain that agents can query. Import from Obsidian, structure with Wiki, retrieve with precision.

Creators — Let agents know your voice, style, past work, and audience without a prompt preamble every time.

Investors — Track deal flow, company updates, and portfolio context across every AI tool you use.

Marketers — Give agents the brand context, past campaigns, and audience knowledge they need to produce on-brand output instantly.

Built By