Persistent long-term
memory for every AI agent
Memobase is a shared memory layer for your entire AI toolchain. Claude, Cursor, and any MCP-compatible agent all read from the same store — secured per user, with sub-second vector search.
No credit card required · Free tier always available
The Problem
Each AI agent lives
in its own silo —
none of them share context
You use Claude for writing, Cursor for coding, and a custom agent for research. Each one knows nothing about what the others have learned about you.
Context stays trapped inside individual tools. There's no shared memory layer — so every agent starts from scratch, and you repeat yourself constantly.
Memobase is the memory layer that sits underneath all of them.
Without Memobase
With Memobase
How It Works
Up and running in minutes
Add the MCP server
Paste one JSON snippet into your Claude Desktop config. Takes 30 seconds — no CLI, no Docker, no fuss.
Authenticate once
Claude prompts you to log in via OAuth. Your memories are cryptographically isolated — no one else can read them.
Start remembering
Claude automatically stores important facts and retrieves relevant context at the start of every conversation.
Features
Built for production from day one
Semantic vector search
Memories are stored as 1536-dimension embeddings using OpenAI's text-embedding-3-small model. Retrieval is by meaning, not keyword — find relevant context even when phrasing differs.
Per-user isolation
PostgreSQL Row-Level Security enforces strict data isolation at the database level. Your memories are cryptographically scoped to your user ID — no sharing, no leakage.
Sub-second retrieval
IVFFlat cosine index delivers approximate nearest-neighbor search in milliseconds, even across millions of stored memories.
MCP-native protocol
Built on the Model Context Protocol — the open standard for AI tool integrations. Works natively with Claude Desktop, Cursor, and any MCP-compatible agent.
Audit-ready
Every memory is timestamped with created_at and updated_at. Attach structured metadata — tags, source, priority — for full traceability.
Usage transparency
Ask Claude "how many memories have I used this month?" any time. The get_memory_usage tool returns your plan, usage count, and reset date in real time.
Pricing
Simple, transparent pricing
Start free and upgrade when you need more. Monthly limits reset on the first of each month.
Free
50 ops / month
Perfect for personal projects and trying Memobase out.
- ✓ store_memory + search_memory (1 op each)
- ✓ forget_memory (always free)
- ✓ Semantic vector search
- ✓ Per-user memory isolation
- ✓ Community support
Pro
2,000 ops / month
For power users and small teams who rely on memory daily.
- ✓ Everything in Free
- ✓ 2,000 operations per month
- ✓ Monthly usage dashboard
- ✓ Email support
- ✓ Priority response
Unlimited
Unlimited ops
No limits. For teams and high-volume agent workflows.
- ✓ Everything in Pro
- ✓ Unlimited operations
- ✓ Dedicated support
- ✓ SLA available
- ✓ Custom onboarding
All plans include: per-user memory isolation · MCP protocol · OpenAI vector embeddings
Quick Setup
30-second install
Add Memobase to Claude Desktop by editing one config file.
Open Claude Desktop settings
Go to Claude → Settings → Developer → Edit Config to open claude_desktop_config.json.
Add the Memobase server
Paste the snippet on the right into your config file under mcpServers.
Restart Claude Desktop
Claude will prompt you to log in. After authentication, the
store_memory,
search_memory, and
forget_memory
tools will be available.
Ask Claude to remember something
Try: "Remember that I prefer TypeScript with strict mode enabled." Claude will store it and retrieve it in every future session.
{
"mcpServers": {
"memobase": {
"command": "npx",
"args": [
"mcp-remote",
"https://mcp.memobase.ai/mcp"
]
}
}
}Requires Node.js 20+ · Works on macOS, Windows, Linux