接入你的智能体
從發布者安裝。不是 PlugMCP 端點。
遠端 URL 還是本機指令？ 兩種安裝方式的差異
來自發布者
Connecting AI agents through shared memory and collaborative intelligence.
Eion is a shared memory storage that provides unified knowledge graph capabilities for multi-agent systems, adapting to different AI deployment scenarios from single LLM applications to complex multi-agency systems.
Shared Memory Storage for Multi-Agent Systems
User ↔ LLM Application → Eion (context storage)
Business Logic ↔ AI Agent → Eion (memory + knowledge graph)
Agent A → context → Agent B → context → Agent C
↓ ↓ ↓
Eion ← shared memory & knowledge → Eion
Agent A ──┐
├── shared live context ← Eion (live sync + notifications)
Agent B ──┤
│
Agent C ──┘
Internal Agency: Agent A ↔ Agent B → Eion ← External Agent C (guest)
↑
(controlled access)
git clone <repo>
cd eion
# Start all required databases (PostgreSQL + Neo4j)
docker-compose up -d
# Verify databases are ready
docker-compose ps
# Enable the pgvector extension (required for embeddings)
docker exec eion_postgres psql -U eion -d eion -c "CREATE EXTENSION IF NOT EXISTS vector;"
# Run main orchestrator migrations (includes sessions table)
docker exec -i eion_postgres psql -U eion -d eion < database_setup.sql
# Create virtual environment
python3 -m venv .venv
source .venv/bin/activate # On Windows: .venv\Scripts\activate
# Install dependencies
pip install -r requirements.txt
# Build the server
go build -o eion-server ./cmd/eion-server
# Run the server
./eion-server
# Check server health
curl http://localhost:8080/health
# Expected response:
# {"status":"healthy","timestamp":"2024-12-19T10:30:00Z","services":{"database":"healthy","embedding":"healthy"}}
Navigate to the Register Console for easy cluster management:
http://localhost:8080/console/
⚠️ Important: Include the trailing slash / in the URL.
Agent Registration Tab:
Monitoring Tab:
Resources Tab:
Register an Agent:
Monitor Activity:
Get Integration Examples:
The console automatically shows:
The console uses Eion's brand colors:
Eion provides a unified API that combines:
all-MiniLM-L6-v2 model (384 dimensions) using sentence-transformers - production-ready embeddingsEion includes a built-in Model Context Protocol (MCP) Server that enables seamless agent integration through standardized tool interfaces.
The MCP server exposes Eion's Session-level API as MCP tools, allowing AI agents to:
Memory Tools (4):
get_memory - Retrieve conversation history with filtersadd_memory - Store new conversation messagessearch_memory - Semantic search in conversation historydelete_memory - Remove conversation dataKnowledge Tools (4):
search_knowledge - Find relevant extracted knowledgecreate_knowledge - Add new knowledge entriesupdate_knowledge - Modify existing knowledgedelete_knowledge - Remove knowledge entriesDirect MCP Client (Recommended):
from mcp import ClientSession, StdioServerParameters
from mcp.client.stdio import stdio_client
# Connect to Eion MCP server
server_params = StdioServerParameters(
command="python",
args=["-m", "internal.mcp.server"],
env={"EION_BASE_URL": "http://localhost:8080"}
)
async with stdio_client(server_params) as (read, write):
async with ClientSession(read, write) as session:
await session.initialize()
# Use tools with agent credentials
result = await session.call_tool("add_memory", {
"session_id": "agent_session_123",
"agent_id": "my_registered_agent",
"user_id": "end_user_456",
"messages": [{"role": "user", "content": "Hello!"}]
})
LangChain Integration:
from langchain_mcp_tools import convert_mcp_to_langchain_tools
mcp_servers = {
"eion": {
"command": "python",
"args": ["-m", "internal.mcp.server"],
"env": {"EION_BASE_URL": "http://localhost:8080"}
}
}
tools, cleanup = await convert_mcp_to_langchain_tools(mcp_servers)
# Use tools with LangChain agents
Claude Desktop Integration:
Add to claude_desktop_config.json:
{
"mcpServers": {
"eion": {
"command": "python",
"args": ["-m", "internal.mcp.server"],
"env": {"EION_BASE_URL": "http://localhost:8080"},
"cwd": "/path/to/eion"
}
}
}
All MCP tools require agent authentication:
session_id - Unique session identifier for the conversationagent_id - Registered agent ID (obtain via Developer API)user_id - End user ID that the agent servesAgents must be registered via the Developer API before using MCP tools:
curl -X POST http://localhost:8080/cluster/v1/agents \
-H "Content-Type: application/json" \
-d '{"name": "My Agent", "capabilities": ["memory", "knowledge"]}'
MCP server settings in eion.yaml:
mcp:
enabled: true # Enable MCP server
port: 8081 # MCP server port
python_path: ".venv/bin/python" # Python executable path
log_level: "INFO" # Logging level
timeout: 30 # Request timeout (seconds)
The MCP server is embedded within the Eion server process and requires no separate deployment.
Create eion.yaml (optional - defaults work out of the box):
common:
http:
host: "0.0.0.0"
port: 8080
postgres:
user: "eion"
password: "eion_pass"
host: "localhost"
port: 5432
database: "eion"
# Neo4j Configuration (Required)
numa:
neo4j:
uri: "bolt://localhost:7687"
username: "neo4j"
password: "password"
database: "neo4j"
# MCP Server Configuration (Optional)
mcp:
enabled: true # Enable MCP server
port: 8081 # MCP server port
python_path: ".venv/bin/python" # Path to Python executable
log_level: "INFO" # MCP server log level
timeout: 30 # Request timeout in seconds
# Register Console Configuration (Optional)
console:
enabled: true # Enable Register Console (default: true)
For production deployments, you may want to customize the database settings in docker-compose.yml or create your own configuration.
Or use the automated setup script:
# One-command setup (includes database startup, Python env, and server build)
./setup.sh
# Then start the server
./eion-server
Shared Memory Storage for Multi-Agent Systems
工具
尚無工具清單。遠端伺服器會在首次健康檢查後取得工具清單。
常見問題
在本頁複製適用於 Cursor、Claude Desktop、Claude Code 或 VS Code 的 stdio 設定。它使用發布者的安裝方式，而不是 PlugMCP 代理伺服器。
這筆記錄未列出必要的環境變數。
Eion 透過 stdio 在本機執行。
上次檢查時間：2026-10-26。 Shared Memory Storage for Multi-Agent Systems