Amber MCP server

Amber

Long-term memory for AI assistants.

Amber is an MCP server that gives any AI assistant persistent, searchable memory across conversations. Your AI remembers preferences, decisions, project context, and personal details - without you doing anything special.

Just talk normally. Amber stores what matters and finds it when relevant.

Quick Install

One command. Works with any MCP-compatible client.

Claude Code / Claude Desktop

claude mcp add --transport http --scope user amber https://mcp.ambermem.com

Cursor

Add to ~/.cursor/mcp.json (or %USERPROFILE%\.cursor\mcp.json on Windows):

{
  "mcpServers": {
    "amber": {
      "url": "https://mcp.ambermem.com"
    }
  }
}

ChatGPT

Settings → Connectors → Create → URL: https://mcp.ambermem.com

Windsurf

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "amber": {
      "serverUrl": "https://mcp.ambermem.com"
    }
  }
}

VS Code (GitHub Copilot)

Add to .vscode/mcp.json:

{
  "servers": {
    "amber": {
      "type": "http",
      "url": "https://mcp.ambermem.com"
    }
  }
}

Any MCP client

URL: https://mcp.ambermem.com | Transport: Streamable HTTP | Auth: OAuth 2.1 (auto-discovered)

How It Works

1. You talk to your AI normally. Amber stores important facts in the background.
2. Next conversation, your AI searches Amber automatically when context would help.
3. Memory improves over time. The more you use it, the better it gets.

No configuration. No tagging. No manual organization.

What Makes Amber Different

Technical Details

• 29 MCP tools (14 memory, 13 account, 2 feedback/notification)
• Hybrid retrieval pipeline: vector search + full-text search + Reciprocal Rank Fusion
• Multi-variant embeddings: every fact is stored with several paraphrases — at least 5, and a store is rejected below that — which is the main thing that makes it findable later
• No generative model reads your memories: your assistant does the chunking, the topics and the phrasings. Memory text is sent only to an embedding endpoint (a vector model, not a chat model), so no chat model is ever shown what you store
• Atomic facts: a conversation becomes individually searchable facts rather than one blob, each with its own subjects, topics and dates
• Temporal search: "what did I say last week?" resolves to a real date range rather than a keyword match
• Automatic topic grouping: memories are grouped by topic, and a search for "work" also finds "career" and "job"
• Async processing: storage completes in the background, never blocking your conversation

Pricing

• 90-day free trial - no charge, cancel anytime
• $2.99/month after trial, via PayPal
• Cancel instantly - ask your AI to cancel, or cancel through PayPal directly
• No lock-in - export all your data as JSON anytime

Privacy

• No email collected
• No marketing, no spam
• Data isolated per user (separate database)
• PayPal handles all payment info
• Full export + account deletion available
• GDPR compliant (data minimization by design)

Architecture

Shortened. The full README is on GitHub.