Long-term memory for AI assistants. Hybrid retrieval, query expansion, auto-topics.
SaferSkills independently audited Amber (MCP Server) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
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 - once it is connected and the two required settings below are on.
Just talk normally. Amber stores what matters and finds it when relevant.
One command. Works with any MCP-compatible client.
claude mcp add --transport http --scope user amber https://mcp.ambermem.comAdd to ~/.cursor/mcp.json (or %USERPROFILE%\.cursor\mcp.json on Windows):
{
"mcpServers": {
"amber": {
"url": "https://mcp.ambermem.com"
}
}
}Settings → Connectors → Create → URL: https://mcp.ambermem.com
Add to ~/.codeium/windsurf/mcp_config.json:
{
"mcpServers": {
"amber": {
"serverUrl": "https://mcp.ambermem.com"
}
}
}Add to .vscode/mcp.json:
{
"servers": {
"amber": {
"type": "http",
"url": "https://mcp.ambermem.com"
}
}
}URL: https://mcp.ambermem.com | Transport: Streamable HTTP | Auth: OAuth 2.1 (auto-discovered)
Adding the URL is not enough on its own. Without these two settings, most assistants use Amber only when you explicitly say "remember this", and their built-in memory quietly takes everything else, so Amber stays nearly empty.
Whenever something worth remembering comes up - a life event, a decision, a preference, something I learned or realized, or anything about who I am - save it to Amber (my memory) without asking. Treat normal conversation as a source of memories, not only when I say "remember this".
Two settings, once. No tagging. No manual organization.
| Feature | Basic memory servers | Amber |
|---|---|---|
| Storage | One embedding per memory | Multiple semantic variants per fact |
| Search | Single vector lookup | Hybrid: vector + keyword + RRF fusion |
| Queries | Exact match only | Many phrasings per fact, matched semantically |
| Input | Stored as-is | Chunked into atomic facts, each independently searchable |
| Topics | Manual tags or none | Auto-grouped, matched semantically at search time |
| Privacy | Varies by server | No generative model reads your memories |
| Time | No temporal awareness | Natural language time parsing ("last week", "3 days ago") |
Amber runs on Cloudflare Workers (zero cold starts, global edge deployment) with Turso databases (one per user, full isolation). Your assistant does the chunking, the topics and the phrasings, so no generative model is ever shown what you store; memory text is sent only to OpenAI's embedding endpoint.
For full technical documentation: ambermem.com/llms.txt
https://mcp.ambermem.comamber_send_feedback_to_developer tool directly.Will it slow my AI down? No. Storage is async (background). Search adds <1 second.
What if Amber shuts down? Export all your data as JSON anytime. Your data is always yours.
Do I need a PayPal account? Currently yes. PayPal handles both identity and billing. More login options coming soon.
Is my data safe? Each user gets a completely isolated database. No data is shared between users. Amber has no access to your PayPal payment details.
Can I self-host? Not currently. Amber is a managed service. We handle the infrastructure, scaling, and LLM costs so you don't have to.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.