The memory layer your AI agents have been missing.
Give your LLM a persistent, structured, relevance-scored memory — that survives every conversation, every session, every project.
LLMs are stateless. Every new conversation starts from zero — no memory of past decisions, preferences, or project context. You repeat yourself constantly. Your AI assistant forgets everything you taught it yesterday.
Synatyx fixes that.
Synatyx is a Context Engine that plugs into any MCP-compatible AI client (Augment Code, Cursor, Claude Desktop, Claude Code) and gives it persistent, structured memory across all your conversations.
flowchart LR
IDE(["🖥️ Your IDE\nAugment / Cursor / Claude"])
MCP["⚙️ Synatyx"]
LLM(["🤖 LLM"])
IDE -->|MCP stdio| MCP
MCP -->|relevant context injected| LLM
subgraph Memory ["4-Layer Memory"]
L1["🔴 Working Memory"]
L2["🟠 Episodic Summaries"]
L3["🟡 Semantic Knowledge"]
L4["🟢 Permanent Rules"]
end
MCP <-->|read / write| Memory
Your AI now remembers what you decided last week, recalls how your codebase is structured, and follows your preferences without being told again.
🧠 Persistent memory across sessions Store facts, decisions, and context once — retrieve them forever. No more repeating yourself.
🎯 Relevance-ranked retrieval Hybrid dense + BM25 + MMR pipeline surfaces the right memories, not just the newest ones.
📦 Multi-project isolation Each project gets its own memory space. Switch projects, switch context — nothing bleeds over.
🔖 Checkpoints that never disappear Pin critical decisions as named snapshots. Deprecate when superseded — never permanently deleted.
✅ Persistent task tracking Tasks survive across sessions. Your AI picks up where it left off.
🤖 Agent skill registry Store, index, and RAG-search agent skill definitions. The right agent for the right task, automatically.
🏭 Production-ready Docker Compose, Alembic migrations, health checks, audit log — ready to deploy.
| Client | Integration |
|---|---|
| Augment Code | MCP stdio |
| Cursor | MCP stdio |
| Claude Desktop | MCP stdio |
| Claude Code | MCP stdio |
| Any MCP client | JSON-RPC 2.0 / stdio |
git clone https://github.com/tanerincode/synatyx.git && cd synatyx
cp .env.example .env # add your EMBEDDING_OPENAI_API_KEY
make # starts everything + tails logs| Doc | What's inside |
|---|---|
| Local Setup | Prerequisites, Docker, IDE config, Makefile reference, troubleshooting |
| MCP Tools Reference | All 18 tools — params, descriptions, examples |
| Architecture | 4-layer memory model, retrieval pipeline, tech stack, project structure |
MIT © Taner Tombas