MCP 服务器

Remembra

Persistent memory layer for AI agents with entity resolution, PII detection, AES-256-GCM encryption at rest, and hybrid search. Self-hosted. 100% on LoCoMo benchmark.

README

<h1 align="center">Remembra</h1>

The memory layer for AI that actually works. Persistent memory with entity resolution, temporal decay, and graph-aware recall. Self-host in minutes. No vendor lock-in.

<a href="https://docs.remembra.dev">Documentation</a> • <a href="https://remembra.dev">Website</a> • <a href="#quick-start">Quick Start</a> • <a href="#why-remembra">Why Remembra?</a> • <a href="https://twitter.com/remembradev">Twitter</a> • <a href="https://discord.gg/Bzv3JshRa3">Discord</a>

🚀 What's New in v0.8.2

🔐 AES-256-GCM Field Encryption — Encrypt memory content at rest with OWASP-compliant key derivation
🛡️ Enterprise Security Suite — PII detection, anomaly monitoring, audit logging
📦 MCP Registry Published — Discoverable as io.github.remembra-ai/remembra in Claude Desktop
⚡ One-Command Quick Start — curl | bash zero-config setup with Ollama embeddings
🔌 Multi-Provider Support — OpenAI, Anthropic Claude, Ollama for embeddings & entity extraction
📊 Usage Warning Banners — API responses include usage thresholds at 60/80/95%

The Problem

Every AI app needs memory. Your chatbot forgets users between sessions. Your agent can't recall decisions from yesterday. Your assistant asks the same questions over and over.

The current solutions suck:

Mem0: $249/mo for graph features, self-hosting docs are trash
Zep: Academic, complex to deploy
Letta: Research-grade, not production-ready
LangChain Memory: Too basic, no persistence

The Solution

from remembra import Memory

memory = Memory(user_id="user_123")

# Store — entities and facts extracted automatically
memory.store("Had a meeting with Sarah from Acme Corp. She prefers email over Slack.")

# Recall — semantic search finds relevant memories
result = memory.recall("How should I contact Sarah?")
print(result.context)
# → "Sarah from Acme Corp prefers email over Slack."

# It knows "Sarah" and "Acme Corp" are entities. It builds relationships.
# It persists across sessions, reboots, context windows. Forever.

⚡ Quick Start (2 Minutes)

One Command Install

curl -sSL https://raw.githubusercontent.com/remembra-ai/remembra/main/quickstart.sh | bash

That's it. Remembra + Qdrant + Ollama start locally. No API keys needed.

Or with Docker Compose directly:

git clone https://github.com/remembra-ai/remembra && cd remembra
docker compose -f docker-compose.quickstart.yml up -d

Try it:

# Store a memory
curl -X POST http://localhost:8787/api/v1/memories/store \
  -H "Content-Type: application/json" \
  -d '{"content": "Alice is CEO of Acme Corp", "user_id": "demo"}'

# Recall it
curl -X POST http://localhost:8787/api/v1/memories/recall \
  -H "Content-Type: application/json" \
  -d '{"query": "Who runs Acme?", "user_id": "demo"}'

Connect to Claude (MCP)

Claude Desktop — add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "remembra": {
      "command": "remembra-mcp",
      "env": {
        "REMEMBRA_URL": "http://localhost:8787",
        "REMEMBRA_USER_ID": "default"
      }
    }
  }
}

Claude Code:

claude mcp add remembra -e REMEMBRA_URL=http://localhost:8787 -- remembra-mcp

Cursor — add to .cursor/mcp.json:

{
  "mcpServers": {
    "remembra": {
      "command": "remembra-mcp",
      "env": {
        "REMEMBRA_URL": "http://localhost:8787"
      }
    }
  }
}

Now ask Claude: "Remember that Alice is CEO of Acme Corp" — then later: "Who runs Acme?"

Python SDK

pip install remembra

from remembra import Memory

memory = Memory(user_id="user_123")
memory.store("Had a meeting with Sarah from Acme Corp. She prefers email over Slack.")
result = memory.recall("How should I contact Sarah?")
print(result.context)  # "Sarah from Acme Corp prefers email over Slack."

TypeScript SDK

npm install remembra

import { Remembra } from 'remembra';

const memory = new Remembra({ url: 'http://localhost:8787' });
await memory.store('User prefers dark mode');
const result = await memory.recall('preferences');

🔥 Why Remembra?

Feature Comparison

Feature	Remembra	Mem0	Zep/Graphiti	Letta	Engram
One-Command Install	✅ `curl \| bash`	✅ pip	✅ pip	⚠️ Complex	✅ brew
Entity Resolution	✅ Free	💰 $249/mo	✅	❌	❌
Conflict Detection	✅ Unique	❌	❌	❌	❌
PII Detection	✅ Built-in	❌	❌	❌	❌
Hybrid Search	✅ BM25+Vector	❌	✅	❌	❌
6 Embedding Providers	✅ Hot-swap	❌ (1-2)	❌ (1)	❌	❌
Plugin System	✅	❌	❌	✅	❌
Sleep-Time Compute	✅	❌	❌	✅	❌
Self-Host + Billing	✅ Stripe	❌	❌	❌	❌
Memory Spaces	✅ Multi-tenant	❌	❌	❌	❌
MCP Server	✅ Native	✅	❌	❌	✅
Pricing	Free / $49 / $99	$19 → $249	$25+	Free	Free
License	MIT	Apache 2.0	Apache 2.0	Apache 2.0	MIT

Core Features

🧠 Smart Extraction — LLM-powered fact extraction from raw text

👥 Entity Resolution — "Adam", "Mr. Smith", "my husband" → same person

⏱️ Temporal Memory — TTL, decay curves, historical queries

🔍 Hybrid Search — Semantic + keyword for accurate recall

🔒 Security — PII detection, anomaly monitoring, audit logs

📊 Dashboard — Visual memory browser, entity graphs, analytics

📊 Benchmark Results

Tested on the LoCoMo benchmark (Snap Research, ACL 2024) — the standard academic benchmark for AI memory systems.

Category	Accuracy	Questions
Single-hop (direct recall)	100%	37
Multi-hop (cross-session reasoning)	100%	32
Temporal (time-based queries)	100%	13
Open-domain (world knowledge + memory)	100%	70
Overall (memory categories)	100%	152

Scored with LLM judge (GPT-4o-mini). Adversarial detection not yet implemented. Run your own: python benchmarks/locomo_runner.py --data /tmp/locomo/data/locomo10.json

📖 Documentation

Resource	Description
Quick Start	Get running in minutes
Python SDK	Full Python reference
TypeScript SDK	JavaScript/TypeScript guide
MCP Server	Tool reference + setup guides for 9 tools
REST API	API reference
Self-Hosting	Docker deployment guide

🛠️ MCP Server

Give any AI coding tool persistent memory with one command. Works with Claude Code, Cursor, VS Code + Copilot, Windsurf, JetBrains, Zed, OpenAI Codex, and any MCP-compatible client.

pip install remembra[mcp]
claude mcp add remembra -e REMEMBRA_URL=http://localhost:8787 -- remembra-mcp

Available Tools:

Tool	Description
`store_memory`	Save facts, decisions, context
`recall_memories`	Semantic search across memories
`forget_memories`	GDPR-compliant deletion
`ingest_conversation`	Auto-extract from chat history
`health_check`	Verify connection

🏗️ Architecture

┌─────────────────────────────────────────────────────────────┐
│                    Your Application                          │
├──────────┬──────────────┬───────────────────────────────────┤
│ Python   │ TypeScript   │ MCP Server (Claude/Cursor)        │
│ SDK      │ SDK          │ remembra-mcp                      │
├──────────┴──────────────┴───────────────────────────────────┤
│                   Remembra REST API                          │
├──────────────┬──────────────┬───────────────┬───────────────┤
│  Extraction  │   Entities   │   Retrieval   │   Security    │
│  (LLM)       │  (Graph)     │ (Hybrid)      │  (PII/Audit)  │
├──────────────┴──────────────┴───────────────┴───────────────┤
│                    Storage Layer                             │
│         Qdrant (vectors) + SQLite (metadata/graph)          │
└─────────────────────────────────────────────────────────────┘

🤝 Contributing

We welcome contributions! See CONTRIBUTING.md for guidelines.

# Clone
git clone https://github.com/remembra-ai/remembra
cd remembra

# Install dev dependencies
pip install -e ".[dev]"

# Run tests
pytest

# Start dev server
remembra-server --reload

📄 License

MIT License — Use it however you want.

⭐ Star History

If Remembra helps you, please star the repo! It helps others discover the project.

Built with ❤️ by <a href="https://dolphytech.com">DolphyTech</a> <a href="https://remembra.dev">remembra.dev</a> • <a href="https://docs.remembra.dev">docs</a> • <a href="https://twitter.com/remembradev">twitter</a> • <a href="https://discord.gg/Bzv3JshRa3">discord</a>