Lets AI assistants evaluate and critique their own outputs using Mandoline's evaluation framework.
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。
claude mcp add mandoline --env MANDOLINE_API_KEY='<MANDOLINE_API_KEY>' -- npx -y mcp-remote https://mandoline.ai/mcp --header 'x-api-key: ${MANDOLINE_API_KEY}'~/.codex/config.toml
[mcp_servers.mandoline]
command = "npx"
args = ["-y", "mcp-remote", "https://mandoline.ai/mcp", "--header", "x-api-key: ${MANDOLINE_API_KEY}"]
env = { MANDOLINE_API_KEY = "<MANDOLINE_API_KEY>" }
# codex mcp add mandoline --env MANDOLINE_API_KEY='<MANDOLINE_API_KEY>' -- npx -y mcp-remote https://mandoline.ai/mcp --header 'x-api-key: ${MANDOLINE_API_KEY}'~/.cursor/mcp.json
{
"mcpServers": {
"mandoline": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://mandoline.ai/mcp",
"--header",
"x-api-key: ${MANDOLINE_API_KEY}"
],
"env": {
"MANDOLINE_API_KEY": "<MANDOLINE_API_KEY>"
}
}
}
}.vscode/mcp.json
{
"servers": {
"mandoline": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://mandoline.ai/mcp",
"--header",
"x-api-key: ${MANDOLINE_API_KEY}"
],
"env": {
"MANDOLINE_API_KEY": "<MANDOLINE_API_KEY>"
}
}
}
}~/.gemini/settings.json
{
"mcpServers": {
"mandoline": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://mandoline.ai/mcp",
"--header",
"x-api-key: ${MANDOLINE_API_KEY}"
],
"env": {
"MANDOLINE_API_KEY": "<MANDOLINE_API_KEY>"
}
}
}
}claude_desktop_config.json
{
"mcpServers": {
"mandoline": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://mandoline.ai/mcp",
"--header",
"x-api-key: ${MANDOLINE_API_KEY}"
],
"env": {
"MANDOLINE_API_KEY": "<MANDOLINE_API_KEY>"
}
}
}
}.trae/mcp.json
{
"mcpServers": {
"mandoline": {
"command": "npx",
"args": [
"-y",
"mcp-remote",
"https://mandoline.ai/mcp",
"--header",
"x-api-key: ${MANDOLINE_API_KEY}"
],
"env": {
"MANDOLINE_API_KEY": "<MANDOLINE_API_KEY>"
}
}
}
}需要的环境变量
MANDOLINE_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。