Store, compress, and retrieve long-term memories with semantic lossless compression. Now with multimodal support for text, image, audio & video. Works across Claude, Cursor, LM Studio, and more.
Works with any AI platform that supports MCP or Python integration
Claude Desktop |
Cursor |
LM Studio |
Cherry Studio |
PyPI Package |
+ Any MCP Client |
🇨🇳 中文 •
🇯🇵 日本語 •
🇰🇷 한국어 •
🇪🇸 Español •
🇫🇷 Français •
🇩🇪 Deutsch •
🇧🇷 Português
🇷🇺 Русский •
🇸🇦 العربية •
🇮🇹 Italiano •
🇻🇳 Tiếng Việt •
🇹🇷 Türkçe
🚀 Quick Start • 🌟 Overview • 📈 Results • 🧠 Omni-SimpleMem • 🧬 EvolveMem • 📦 Installation • 🔌 MCP Server • 📝 Citation
- [05/14/2026] 🧬 EvolveMem (v3.0) — Self-Evolving Memory via AutoResearch! The retrieval infrastructure itself now self-evolves through LLM-driven closed-loop diagnosis. On LoCoMo, EvolveMem outperforms the strongest baseline by +25.7% relative; on MemBench, by +18.9% relative. The system discovers entirely new retrieval dimensions not present in the original design. View EvolveMem →
- [04/02/2026] 🧠 Omni-SimpleMem (v2.0) — Multimodal Memory is Here! SimpleMem now supports text, image, audio & video memory. Achieving new SOTA on LoCoMo (F1=0.613, +47%) and Mem-Gallery (F1=0.810, +51%) over previous best. View Omni-SimpleMem →
- [02/09/2026] 🚀 Cross-Session Memory — Outperforming Claude-Mem by 64%! View Cross-Session Documentation →
- [01/20/2026] 📦 SimpleMem is now available on PyPI! Install via
pip install simplemem. View Package Usage Guide → - [01/14/2026] 🎉 SimpleMem MCP Server is LIVE! Cloud-hosted at mcp.simplemem.cloud. View MCP Documentation →
- [01/05/2026] SimpleMem paper was released on arXiv!
- 🚀 Quick Start
- 🌟 Overview
- 📈 Results
- 📝 SimpleMem: Text Memory
- 🧠 Omni-SimpleMem: Multimodal Memory
- 🧬 EvolveMem: Self-Evolving Memory
- 📦 Installation
- 🐳 Docker
- 🔌 MCP Server
- 📊 Evaluation
- 📝 Citation
At a high level, SimpleMem works as a long-term memory system for LLM-based agents. The workflow consists of three simple steps:
- Store information – Dialogues or facts are processed and converted into structured, atomic memories.
- Index memory – Stored memories are organized using semantic embeddings and structured metadata.
- Retrieve relevant memory – When a query is made, SimpleMem retrieves the most relevant stored information based on meaning rather than keywords.
This design allows LLM agents to maintain context, recall past information efficiently, and avoid repeatedly processing redundant history.
SimpleMem provides a unified entry point via simplemem_router. The default mode="auto" automatically detects which backend to use based on what you call — no manual configuration needed:
import simplemem_router as simplemem
mem = simplemem.create() # mode="auto" — backend chosen by first callThe first method you call determines the backend:
| First call | Backend selected | Why |
|---|---|---|
add_dialogue() |
Text (SimpleMem) | Dialogue-based API → text mode |
add_text() / add_image() / add_audio() / add_video() |
Omni (Omni-SimpleMem) | Multimodal API → omni mode |
|
📝 Auto → Text (pure text input) import simplemem_router as simplemem
mem = simplemem.create() # auto mode
# add_dialogue() → text backend auto-selected
mem.add_dialogue(
"Alice",
"Bob, let's meet at Starbucks tomorrow at 2pm",
"2025-11-15T14:30:00",
)
mem.add_dialogue(
"Bob",
"Sure, I'll bring the market analysis report",
"2025-11-15T14:31:00",
)
mem.finalize()
answer = mem.ask("When and where will Alice and Bob meet?")
# → "16 November 2025 at 2:00 PM at Starbucks" |
🧠 Auto → Omni (multimodal input) import simplemem_router as simplemem
mem = simplemem.create() # auto mode
# add_image() → omni backend auto-selected
mem.add_text(
"User loves hiking in the Rocky Mountains.",
tags=["session_id:D1"],
)
mem.add_image("photo.jpg", tags=["session_id:D1"])
mem.add_audio("voice_note.wav", tags=["session_id:D1"])
result = mem.query("What does the user enjoy?", top_k=5)
for item in result.items:
print(item["summary"])
mem.close() |
💡 Tip: Auto mode picks the lightest backend that fits your data. You can still use
mode="text"ormode="omni"explicitly if you prefer.
For large-scale dialogue processing, enable parallel mode:
import simplemem_router as simplemem
mem = simplemem.create(
mode="text",
clear_db=True,
enable_parallel_processing=True, # ⚡ Parallel memory building
max_parallel_workers=8,
enable_parallel_retrieval=True, # 🔍 Parallel query execution
max_retrieval_workers=4
)💡 Pro Tip: Parallel processing significantly reduces latency for batch operations!
SimpleMem is a family of efficient memory frameworks — SimpleMem for text and Omni-SimpleMem for multimodal (text, image, audio, video) — based on semantic lossless compression that addresses the fundamental challenge of efficient long-term memory for LLM agents. Unlike existing systems that either passively accumulate redundant context or rely on expensive iterative reasoning loops, SimpleMem maximizes information density and token utilization through a three-stage pipeline:
|
Semantic Structured Compression Distills unstructured interactions into compact, multi-view indexed memory units |
Online Semantic Synthesis Intra-session process that instantly integrates related context into unified abstract representations to eliminate redundancy |
Intent-Aware Retrieval Planning Infers search intent to dynamically determine retrieval scope and construct precise context efficiently |
For multimodal memory, see Omni-SimpleMem below.
SimpleMem achieves superior F1 score (43.24%) with minimal token cost (~550), occupying the ideal top-left position.
Speed Comparison Demo
SimpleMem vs. Baseline: Real-time speed comparison demonstration
LoCoMo-10 Benchmark Results (GPT-4.1-mini)
| Model | ⏱️ Construction Time | 🔎 Retrieval Time | ⚡ Total Time | 🎯 Average F1 |
|---|---|---|---|---|
| A-Mem | 5140.5s | 796.7s | 5937.2s | 32.58% |
| LightMem | 97.8s | 577.1s | 675.9s | 24.63% |
| Mem0 | 1350.9s | 583.4s | 1934.3s | 34.20% |
| SimpleMem ⭐ | 92.6s | 388.3s | 480.9s | 43.24% |
🏆 Cross-Session Memory Comparison
| System | LoCoMo Score | vs SimpleMem |
|---|---|---|
| SimpleMem | 48 | — |
| Claude-Mem | 29.3 | +64% |
🔬 High-Capability Models (GPT-4.1-mini)
| Task Type | SimpleMem F1 | Mem0 F1 | Improvement |
|---|---|---|---|
| MultiHop | 43.46% | 30.14% | +43.8% |
| Temporal | 58.62% | 48.91% | +19.9% |
| SingleHop | 51.12% | 41.3% | +23.8% |
⚙️ Efficient Models (Qwen2.5-1.5B)
| Metric | SimpleMem | Mem0 | Notes |
|---|---|---|---|
| Average F1 | 25.23% | 23.77% | Competitive with 99× smaller model |
| 🏆 0.543 F1 LoCoMo GPT-4o (+25.7% over SimpleMem) |
🏆 0.572 F1 LoCoMo GPT-5.1 (+36.8% over SimpleMem) |
🏆 71.4% Acc MemBench (+18.9% over best baseline) |
🧬 Self-evolving 7 autonomous rounds |
| 🏆 0.613 F1 LoCoMo (+47% over prev. SOTA) |
🏆 0.810 F1 Mem-Gallery (+51% over prev. SOTA) |
⚡ 3.5x faster retrieval throughput |
🧠 4 modalities Text · Image · Audio · Video |
SimpleMem applies an implicit semantic density gating mechanism integrated into the LLM generation process to filter redundant interaction content. The system reformulates raw dialogue streams into compact memory units — self-contained facts with resolved coreferences and absolute timestamps. Each unit is indexed through three complementary representations for flexible retrieval:
| 🔍 Layer | 📊 Type | 🎯 Purpose | 🛠️ Implementation |
|---|---|---|---|
| Semantic | Dense | Conceptual similarity | Vector embeddings (1024-d) |
| Lexical | Sparse | Exact term matching | BM25-style keyword index |
| Symbolic | Metadata | Structured filtering | Timestamps, entities, persons |
✨ Example Transformation:
- Input: "He'll meet Bob tomorrow at 2pm" [❌ relative, ambiguous]
+ Output: "Alice will meet Bob at Starbucks on 2025-11-16T14:00:00" [✅ absolute, atomic]Unlike traditional systems that rely on asynchronous background maintenance, SimpleMem performs synthesis on-the-fly during the write phase. Related memory units are synthesized into higher-level abstract representations within the current session scope, allowing repetitive or structurally similar experiences to be denoised and compressed immediately.
✨ Example Synthesis:
- Fragment 1: "User wants coffee"
- Fragment 2: "User prefers oat milk"
- Fragment 3: "User likes it hot"
+ Consolidated: "User prefers hot coffee with oat milk"This proactive synthesis ensures the memory topology remains compact and free of redundant fragmentation.
Instead of fixed-depth retrieval, SimpleMem leverages the reasoning capabilities of the LLM to generate a comprehensive retrieval plan. Given a query, the planning module infers latent search intent to dynamically determine retrieval scope and depth:
The system then executes parallel multi-view retrieval across semantic, lexical, and symbolic indexes, and merges results through ID-based deduplication:
|
🔹 Simple Queries
|
🔸 Complex Queries
|
📈 Result: 43.24% F1 score with 30× fewer tokens than full-context methods.
Omni-SimpleMem extends SimpleMem to unified multimodal memory — supporting text, image, audio, and video experiences with state-of-the-art accuracy across all five LLM backbones tested.
Built on three principles: Selective Ingestion (entropy-driven filtering for each modality), Progressive Retrieval (hybrid FAISS + BM25 search with pyramid token-budget expansion), and Knowledge Graph Augmentation (multi-hop cross-modal reasoning).
📖 Full documentation, benchmarks, and architecture details: Omni-SimpleMem →
EvolveMem (v3.0) makes the retrieval infrastructure itself a first-class optimization target. While SimpleMem and Omni-SimpleMem keep retrieval configurations frozen, EvolveMem autonomously evolves its retrieval policy through an LLM-driven closed-loop:
Evaluate → Diagnose failures → Propose config changes → Guard against regression → Repeat
This self-evolution constitutes an AutoResearch process: the system conducts iterative research cycles on its own architecture, discovering new retrieval dimensions (query decomposition, entity-swap, answer verification) that were not in the original design.
| Benchmark | Backbone | EvolveMem | Best Baseline | Relative Gain |
|---|---|---|---|---|
| LoCoMo (F1) | GPT-4o | 0.543 | 0.432 (SimpleMem) | +25.7% |
| LoCoMo (F1) | GPT-5.1 | 0.572 | 0.418 (SimpleMem) | +36.8% |
| MemBench (Acc) | GPT-4o | 67.9% | 57.1% | +18.9% |
| MemBench (Acc) | GPT-5.1 | 71.4% | 64.3% | +11.0% |
📖 Full documentation, architecture, and usage: EvolveMem →
- Ensure you are using Python 3.10 in your active environment, not just installed globally.
- An OpenAI-compatible API key must be configured before running any memory construction or retrieval, otherwise initialization may fail.
- When using non-OpenAI providers (e.g., Qwen or Azure OpenAI), verify both the model name and
OPENAI_BASE_URLinconfig.py. - For large dialogue datasets, enabling parallel processing can significantly reduce memory construction time.
- 🐍 Python 3.10
- 🔑 OpenAI-compatible API (OpenAI, Qwen, Azure OpenAI, etc.)
# 📥 Clone repository
git clone https://github.com/aiming-lab/SimpleMem.git
cd SimpleMem
# 📦 Install dependencies
pip install -r requirements.txt
# ⚙️ Configure API settings
cp config.py.example config.py
# Edit config.py with your API key and preferences# config.py
OPENAI_API_KEY = "your-api-key"
OPENAI_BASE_URL = None # or custom endpoint for Qwen/Azure
LLM_MODEL = "gpt-4.1-mini"
EMBEDDING_MODEL = "Qwen/Qwen3-Embedding-0.6B" # State-of-the-art retrievalThe MCP Server can be run in Docker for a consistent, isolated environment. Data (LanceDB and user DB) is persisted in a host volume.
- Docker and Docker Compose
# From the repository root
docker compose up -d- Web UI: http://localhost:8000/
- REST API: http://localhost:8000/api/
- MCP (SSE): http://localhost:8000/mcp/sse?token=<TOKEN>
Data is stored in ./data on the host (created automatically).
- Copy the environment template and edit it:
cp .env.example .env # Edit .env: set JWT_SECRET_KEY, ENCRYPTION_KEY, LLM_PROVIDER, model URLs, etc. - Run with the env file:
docker compose --env-file .env up -d
When LLM_PROVIDER=ollama and Ollama runs on your machine (not in Docker), set in .env:
LLM_PROVIDER=ollama
OLLAMA_BASE_URL=http://host.docker.internal:11434/v1On Linux, host.docker.internal is enabled automatically via the Compose file.
docker compose logs -f simplemem # Follow logs
docker compose down # Stop and remove containers📖 For self-hosting the MCP server (Docker or bare metal), see MCP Documentation.
SimpleMem is available as a cloud-hosted memory service via the Model Context Protocol (MCP), enabling seamless integration with AI assistants like Claude Desktop, Cursor, and other MCP-compatible clients.
🌐 Cloud Service: mcp.simplemem.cloud — or self-host the MCP server locally using Docker.
| Feature | Description |
|---|---|
| Streamable HTTP | MCP 2025-03-26 protocol with JSON-RPC 2.0 |
| Multi-tenant Isolation | Per-user data tables with token authentication |
| Hybrid Retrieval | Semantic search + keyword matching + metadata filtering |
| Production Optimized | Faster response times with OpenRouter integration |
{
"mcpServers": {
"simplemem": {
"url": "https://mcp.simplemem.cloud/mcp",
"headers": {
"Authorization": "Bearer YOUR_TOKEN"
}
}
}
}📖 For detailed setup instructions and self-hosting guide, see MCP Documentation
# 🎯 Full LoCoMo benchmark
python test_locomo10.py
# 📉 Subset evaluation (5 samples)
python test_locomo10.py --num-samples 5
# 💾 Custom output file
python test_locomo10.py --result-file my_results.jsonUse the exact configurations in config.py:
- 🚀 High-capability: GPT-4.1-mini, Qwen3-Plus
- ⚙️ Efficient: Qwen2.5-1.5B, Qwen2.5-3B
- 🔍 Embedding: Qwen3-Embedding-0.6B (1024-d)
If you use SimpleMem in your research, please cite:
@article{simplemem2026,
title={SimpleMem: Efficient Lifelong Memory for LLM Agents},
author={Liu, Jiaqi and Su, Yaofeng and Xia, Peng and Zhou, Yiyang and Han, Siwei and Zheng, Zeyu and Xie, Cihang and Ding, Mingyu and Yao, Huaxiu},
journal={arXiv preprint arXiv:2601.02553},
year={2026},
url={https://arxiv.org/abs/2601.02553}
}@article{evolvemem2026,
title={EvolveMem: Self-Evolving Memory Architecture via AutoResearch for LLM Agents},
author={Liu, Jiaqi and Ye, Xinyu and Xia, Peng and Zheng, Zeyu and Xie, Cihang and Ding, Mingyu and Yao, Huaxiu},
journal={arXiv preprint arXiv:2605.13941},
year={2026},
url={https://arxiv.org/abs/2605.13941}
}@article{omnisimplemem2026,
title = {Omni-SimpleMem: Autoresearch-Guided Discovery of Lifelong Multimodal Agent Memory},
author = {Liu, Jiaqi and Ling, Zipeng and Qiu, Shi and Liu, Yanqing and Han, Siwei and Xia, Peng and Tu, Haoqin and Zheng, Zeyu and Xie, Cihang and Fleming, Charles and Ding, Mingyu and Yao, Huaxiu},
journal = {arXiv preprint arXiv:2604.01007},
year = {2026},
}This project is licensed under the MIT License - see the LICENSE file for details.
We would like to thank the following projects and teams:
- 🔍 Embedding Model: Qwen3-Embedding - State-of-the-art retrieval performance
- 🗄️ Vector Database: LanceDB - High-performance columnar storage
- 📊 Benchmark: LoCoMo - Long-context memory evaluation framework