Agenten stärken
Installation vom Herausgeber. Kein PlugMCP-Endpunkt.
Host a copy with our keys and logs.
Noch kein paketiertes Release in der Registry. Folgen Sie der README des Repositorys.
Remote-URL oder ein Befehl auf Ihrem Rechner? Worin sich die beiden Installationen unterscheiden
Auf PlugMCP ausführen
A new server you create. It is not this listing's official endpoint. Use the publisher install when you want their URL and their credentials.
StartenVom Herausgeber
AI agents are getting smarter every day — but they still suffer from a fundamental limitation: they forget.
Different agents working on the same problem rarely share context, history, or decisions. One agent discovers a breaking API change, another reviews code, a third generates documentation — and none of them know what the others have already learned.
MAGI — Multi-Agent Graph Intelligence. Universal memory server for AI agents. MCP + gRPC + REST API with semantic search, knowledge graph, and behavioral pattern analysis.
Switch machines and your agent has amnesia. Switch providers and your context is gone. Every session starts cold.
In short: plenty to panic about.
Don't.
MAGI is a universal memory server for AI agents. It acts as a shared, persistent brain that any agent can read from and write to — regardless of framework, language, or orchestration model.
Host it on your own hardware, point every agent at the same server, and your context persists across sessions, machines, and providers. No vendor lock-in, no rebuilds, no cold starts.
MAGI exposes memory through multiple protocols — MCP (stdio), gRPC, REST API, and a Web UI — making it easy to integrate into existing workflows without forcing a specific toolchain.
MAGI isn't just another vector store or RAG backend.
Every memory write is committed to a Git repository — full history, diffs, rollback, auditable changes. The database is a derived index, not the source of truth.
Dedicated Writer, Reader, Index, and Coordinator node pools with session affinity for read-your-writes consistency. Supports both embedded and multi-node deployments.
Combines vector embeddings and BM25 for accurate recall across structured and unstructured memories.
Automatically links memories into a knowledge graph and surfaces behavioral patterns — preferences, habits, recurring decisions — across agents.
202 Accepted in under 10ms/readyz, /livez, expanded /health for KubernetesMAGI is fully self-hosted and supports multiple backends:
No cloud lock-in. No hosted dependency. Your data stays on your infrastructure.
Works with — or without — popular orchestrators like Openclaw, LangChain, CrewAI, or custom pipelines.
git clone https://github.com/j33pguy/magi.git
cd magi
docker compose up -d
curl http://localhost:8302/health
magi mcp-config
export MAGI_HTTP_URL=http://localhost:8302
export MAGI_API_TOKEN=your-token
# Agent A stores a decision
curl -X POST "$MAGI_HTTP_URL/remember" \
-H "Authorization: Bearer $MAGI_API_TOKEN" \
-d '{"content":"API v3 deprecates /users","type":"decision","speaker":"agent-a"}'
# Agent B stores a lesson
curl -X POST "$MAGI_HTTP_URL/remember" \
-H "Authorization: Bearer $MAGI_API_TOKEN" \
-d '{"content":"Migrate clients before Q4","type":"lesson","speaker":"agent-b"}'
# Any agent recalls shared context
curl -X POST "$MAGI_HTTP_URL/recall" \
-H "Authorization: Bearer $MAGI_API_TOKEN" \
-d '{"query":"API changes","top_k":5}'
graph LR
A[Agents] -->|MCP / gRPC / REST| M[MAGI]
U[Web UI] --> M
M --> S[Storage Backends]
M --> G[Git History]
Getting Started · Architecture · MCP Tools · REST API · Multi-Agent Setup · Knowledge Graph · Deployment · Config · FAQ
This project is dedicated to Mary Margaret — a dear friend who believed that the things worth remembering are the things that connect us. MAGI carries her spirit: nothing important should ever be forgotten.
MAGI is not production-ready yet. It is useful today and improving fast, but expect breaking changes, rough edges, and the occasional surprise until a stable release is tagged. Back up your data, test in your own environment, and plan for things to break. Bring a towel. SQLite is the most-tested backend; PostgreSQL, MySQL/MariaDB, and SQL Server backends exist but see less CI coverage. The distributed node mesh is architected but currently ships as a single-node embedded process. Git-backed history is optional and off by default.
Elastic License 2.0 (ELv2) — free to use, modify, and self-host. Cannot be offered as a managed/hosted service without a commercial license from the author.
MAGI — Multi-Agent Graph Intelligence. Universal memory server for AI agents. MCP + gRPC + REST API with semantic search, knowledge graph, and behavioral pattern analysis.
Tools
Noch keine Tool-Liste. Remote-Server erhalten eine nach dem ersten Health-Check.
FAQ
Kopieren Sie auf dieser Seite die stdio-Konfiguration für Cursor, Claude Desktop, Claude Code oder VS Code. Sie nutzt die Installation des Herausgebers, keinen PlugMCP-Proxy.
In diesem Eintrag ist keine erforderliche Umgebungsvariable aufgeführt.
Magi läuft lokal über stdio.
Zuletzt geprüft: 2026-10-28. MAGI — Multi-Agent Graph Intelligence. Universal memory server for AI agents. MCP + gRPC + REST API with semantic search, knowledge graph, and behavioral pattern analysis.
