Persistent memory for AI agents via MCP — entity/relation knowledge graph, works with Claude Desktop, Cursor, and Codex CLI
接入配置
选择客户端，复制配置，再按项目文档补齐环境变量。
本地运行
claude mcp add spomory --env LLM_API_KEY='<LLM_API_KEY>' -- uvx spomory~/.codex/config.toml
[mcp_servers.spomory]
command = "uvx"
args = ["spomory"]
env = { LLM_API_KEY = "<LLM_API_KEY>" }
# codex mcp add spomory --env LLM_API_KEY='<LLM_API_KEY>' -- uvx spomory~/.cursor/mcp.json
{
"mcpServers": {
"spomory": {
"command": "uvx",
"args": [
"spomory"
],
"env": {
"LLM_API_KEY": "<LLM_API_KEY>"
}
}
}
}.vscode/mcp.json
{
"servers": {
"spomory": {
"type": "stdio",
"command": "uvx",
"args": [
"spomory"
],
"env": {
"LLM_API_KEY": "<LLM_API_KEY>"
}
}
}
}~/.gemini/settings.json
{
"mcpServers": {
"spomory": {
"command": "uvx",
"args": [
"spomory"
],
"env": {
"LLM_API_KEY": "<LLM_API_KEY>"
}
}
}
}claude_desktop_config.json
{
"mcpServers": {
"spomory": {
"command": "uvx",
"args": [
"spomory"
],
"env": {
"LLM_API_KEY": "<LLM_API_KEY>"
}
}
}
}.trae/mcp.json
{
"mcpServers": {
"spomory": {
"command": "uvx",
"args": [
"spomory"
],
"env": {
"LLM_API_KEY": "<LLM_API_KEY>"
}
}
}
}远程托管
claude mcp add --transport http spomory https://api.yliuai.com/mcp~/.codex/config.toml
[mcp_servers.spomory]
url = "https://api.yliuai.com/mcp"
# codex mcp add spomory --url https://api.yliuai.com/mcp~/.cursor/mcp.json
{
"mcpServers": {
"spomory": {
"url": "https://api.yliuai.com/mcp"
}
}
}.vscode/mcp.json
{
"servers": {
"spomory": {
"type": "http",
"url": "https://api.yliuai.com/mcp"
}
}
}~/.gemini/settings.json
{
"mcpServers": {
"spomory": {
"httpUrl": "https://api.yliuai.com/mcp"
}
}
}claude_desktop_config.json · Customize › Connectors › Add custom connector
{
"mcpServers": {
"spomory": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://api.yliuai.com/mcp"
]
}
}
}.trae/mcp.json
{
"mcpServers": {
"spomory": {
"url": "https://api.yliuai.com/mcp"
}
}
}需要的环境变量
LLM_API_KEY
把占位值换成你自己的密钥；不要把密钥提交到代码仓库。