AmpSAFE
AMP: The Agent Memory Protocol — Open source, MCP-native memory server for AI agents. Give your LLMs a hippocampus.
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
The Open Standard for Agentic Memory.
[](https://opensource.org/licenses/MIT) [](https://www.python.org/downloads/) [](https://modelcontextprotocol.io/introduction) [](https://pypi.org/project/amp-memory/)
The (Short) Story
I was tired of building AI agents that forgot everything the moment I closed the terminal.
RAG (Retrieval Augmented Generation) is great for documents, but terrible for experience. It chunks text blindly, losing the narrative. When I asked my agents "Why did we decide this yesterday?", they gave me hallucinated nonsense.
So I built AMP. It's not just a database; it's a Hippocampus for your agents. It mimics the human brain's distinction between Working Memory (Short-term context) and episodic Long-Term Memory, giving your agents a continuous, evolving sense of self.
Why developers are switching to AMP?
🌌 Galaxy View (Visualization)
Don't just guess what your agent knows. See it. AMP comes with a local dashboard. Watch memories form constellations in real-time. Nodes cluster by semantic meaning—if two ideas are related, they physically move together.
🕸️ Force Mode (Physics)
Toggle to Force Mode to see the topological connections between your memories. It uses a physics simulation (D3.js) to show you how different memory clusters are "pulled" together by shared context.
🔍 Semantic Query
Stop guessing keywords. Query your agent's memory using natural language. I built a dedicated interface that not only finds relevant memories but shows you the Relevance Score (0-100%) so you know exactly why a memory was retrieved.
![Semantic
b231e26d5286OBSERVED · 2026-10-08Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.
claude mcp add amp-memory --env GEMINI_API_KEY=${GEMINI_API_KEY} --env HUGGINGFACE_TOKENIZER=${HUGGINGFACE_TOKENIZER} --env LLM_API_KEY=${LLM_API_KEY} --env MEM0_API_KEY=${MEM0_API_KEY} -- uvx amp-memory{
"mcpServers": {
"amp-memory": {
"command": "uvx",
"args": [
"amp-memory"
],
"env": {
"GEMINI_API_KEY": "${GEMINI_API_KEY}",
"HUGGINGFACE_TOKENIZER": "${HUGGINGFACE_TOKENIZER}",
"LLM_API_KEY": "${LLM_API_KEY}",
"MEM0_API_KEY": "${MEM0_API_KEY}"
}
}
}
}Exposed tools (4)
3 read · 1 write · 0 destructive.
| Tool | Risk | Description |
|---|---|---|
add_memory | write | |
consolidate_memories | read | |
forget_memory | read | |
search_memory | read |
Trust audit
SAFEgrade B · trust 89/100 Nothing in the source contradicts what it says it does. Grade A is reserved for packages that have also passed the behavioural sandbox.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | PASS |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- none-observed
- Network
- none-observed
- Shell
- none-observed
- Dependencies
- pinned
- Secrets in source
- none-found
Findings (0)
No findings outside the package's declared scope.
Gates applied: no_behavioural_pass.
b231e26d5286full audit observations/trust-audit/mcp-server/akshayaggarwal99__amp.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-08 | b231e26d5286 | SAFE | B | 89 | first audit |
Questions
What is the Amp MCP server?
AMP: The Agent Memory Protocol — Open source, MCP-native memory server for AI agents. Give your LLMs a hippocampus.
What tools does Amp expose?
4 in total: 3 read-only, 1 that write, and 0 that can delete or overwrite. Every one is listed on this page with its risk.
Is Amp safe to connect to an agent?
The audit found nothing in the source that contradicts what it says it does, and graded it B (89/100). Grade A is held back for packages that have also passed a sandboxed behavioural run, which is why a clean server reads B.
What credentials does Amp need?
It reads GEMINI_API_KEY, HUGGINGFACE_TOKENIZER, LLM_API_KEY, MEM0_API_KEY and OPENAI_API_KEY from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How does Amp run?
It speaks sse, stdio and streamable-http, so it runs as a local process your client starts. It is published on PyPI as amp-memory.
How current is this page?
The grade is for one exact copy of the source (b231e26d5286), read on 2026-10-08. The repository is watched and re-audited when it changes.