Openai MCP Lunarcrush

openai mcp lunarcrush

目标

演示如何将OpenAI代理SDK连接到LunarCrush MCP服务器，并通过自动代理运行显示加密货币情绪洞察。

先决条件

• LunarCrush帐户和MCP访问权限

- https://lunarcrush.com - https://docs.lunarcrush.com/reference/mcp-getting-started

• 活动OpenAI API订阅（付费即用或企业版）

- https://platform.openai.com/docs/guides/billing/metered-billing

依赖项

• OpenAI代理SDK：https://github.com/openai/openai-agents-python

如何跑步

1. 如果你还没有安装Poetry 2.x。
2. 安装项目依赖关系： poetry install
3. 复制 .env.example 到 .env.local 并填写所需值：

- OPENAI_API_KEY - MCP_SERVER_URL

1. （可选）将任何其他机密添加到 .env.local；脚本也将从 .env 如果存在。
2. 执行代理： poetry run python lunarcrush-openai-mcp-client.py

样本输入和输出

该脚本发送一个提示，要求获得社交动力的顶级加密货币。

$ poetry run python lunarcrush-openai-mcp-client.py

--- Attempt 1/1 ---

=== LLM SUMMARY ===
Top tokens by rising social sentiment include BTC, ETH, and SOL...

=== PARSED METRICS ===
Sentiment: 0.62, Galaxy Score: 68.5, AltRank: 33
Sentiment: 0.58, Galaxy Score: 70.1, AltRank: 41

Query time: 9.42 s

流程图

sequenceDiagram
    participant Client as CLI Client
    participant Agent as OpenAI Agent
    participant MCP as LunarCrush MCP

    Client->>Agent: Execute lunarcrush-openai-mcp-client.py
    Agent->>Agent: Load environment & configure API keys
    Agent->>MCP: Streamable HTTP request (cryptocurrency metrics)
    MCP-->>Agent: MCP response with sentiment payloads
    Agent-->>Client: Final report & parsed metrics

杰出的

• 默认情况下，MCP服务器仍然强制执行5秒的请求超时；延长超时时间 extend_mcp_timeout.py 是等待本机服务器端配置时的临时解决方法。