OpenCode CLI 多专家设置需要一个共享的网络搜索层，以便每个专家代理都能看到相同的新基线。 Scavio 的托管 MCP（位于“https://mcp.scavio.dev/mcp”）只需插入一次即可向每位专家公开搜索。本教程涵盖配置、专家提示设置和路由模式。
前置条件
- OpenCode CLI 已安装
- Scavio API 密钥
操作指南
步骤 1: 将 Scavio MCP 添加到 OpenCode 配置
单一端点、单一 API 密钥。
JSON
// ~/.opencode/config.json
{
"mcpServers": {
"scavio": {
"url": "https://mcp.scavio.dev/mcp",
"apiKey": "${SCAVIO_API_KEY}"
}
}
}步骤 2: 定义专家提示
每个专家角色一个 Markdown 文件。
# experts/seo-expert.md
You are an SEO expert. Always search Scavio for current SERP data before answering.
# experts/code-expert.md
You are a code expert. Search Scavio for current docs and GitHub issues.步骤 3: 布线路由
OpenCode 将用户查询路由给合适的专家。
# experts/router.md
Classify the query: SEO -> seo-expert, code -> code-expert, default -> generalist.步骤 4: 使用 SEO 查询进行测试
验证使用正确的查询调用 Scadio。
Bash
$ opencode 'rank top 5 cars 2026'
# router -> seo-expert
# seo-expert calls Scavio /api/v1/google
# returns SERP plus AI Overview citations步骤 5: 使用代码查询进行测试
验证代码专家通过 SERP 拉取 GitHub。
Bash
$ opencode 'find async-await issues in tokio repo'
# router -> code-expert
# code-expert calls Scavio with site:github.com/tokio-rs filterPython 示例
Python
# OpenCode is a CLI; use Python to orchestrate experts directly via Scavio
import os, requests
API_KEY = os.environ['SCAVIO_API_KEY']
EXPERTS = {'seo': 'site: in any', 'code': 'site:github.com'}
def expert_search(role, q):
scope = EXPERTS.get(role, '')
r = requests.post('https://api.scavio.dev/api/v1/google',
headers={'x-api-key': API_KEY},
json={'query': f'{q} {scope}'}).json()
return r.get('organic_results', [])[:10]JavaScript 示例
JavaScript
const API_KEY = process.env.SCAVIO_API_KEY;
const EXPERTS = { seo: '', code: 'site:github.com' };
export async function expertSearch(role, q) {
const scope = EXPERTS[role] || '';
const r = await fetch('https://api.scavio.dev/api/v1/google', { method:'POST', headers:{'x-api-key':API_KEY,'Content-Type':'application/json'}, body: JSON.stringify({ query: `${q} ${scope}` }) });
return (await r.json()).organic_results || [];
}预期输出
JSON
OpenCode CLI routes user queries to the right expert; each expert calls Scavio MCP for fresh data. Same fresh baseline across all experts.