Amber MCP server
{
"mcpServers": {
"ambermem": {
"type": "http",
"url": "https://mcp.ambermem.com"
}
}
}{
"mcpServers": {
"ambermem": {
"type": "http",
"url": "https://mcp.ambermem.com"
}
}
}{
"mcpServers": {
"ambermem": {
"type": "http",
"url": "https://mcp.ambermem.com"
}
}
}{
"servers": {
"ambermem": {
"type": "http",
"url": "https://mcp.ambermem.com"
}
}
}{
"mcpServers": {
"ambermem": {
"type": "http",
"url": "https://mcp.ambermem.com"
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsonversion 1.1.1295
What it does
Amber is an MCP server that provides persistent, searchable memory for AI assistants across conversations. It stores facts, preferences, and context automatically and retrieves relevant memories during new conversations without manual tagging or configuration.
- Maintain continuity of project context and decisions across multiple conversations with your AI assistant.
- Retrieve personal preferences and past discussions by natural language queries like 'what did I say last week about this?'
- Let your AI automatically find relevant prior context when answering follow-up questions in new sessions.
Tools this server exposes
| Tool | Does |
|---|---|
| amber_store_memory | Store facts and context from the conversation as searchable memories with automatic chunking and semantic variants. |
| amber_search | Search stored memories using hybrid retrieval combining vector search, full-text search, and reciprocal rank fusion. |
| amber_list_memories | Retrieve a list of stored memories, optionally filtered by topic or date range. |
| amber_delete_memory | Remove a specific memory from storage. |
| amber_export_data | Export all stored memories and account data as JSON. |
| amber_delete_account | Permanently delete your account and all associated data. |
| amber_send_feedback_to_developer | Submit feedback or report problems directly to the Amber development team. |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.