The Problem
A "Google AI Mode MCP" is the wrong thing to chase. The handful of GitHub MCP servers that promise Google's AI Overview or AI Mode text all scrape the results page, which breaks the moment Google changes its HTML, gets your IP banned, and violates the ToS. Most of those repos go unmaintained within months. What a coding agent actually needs isn't Google's black-box AI summary. It needs the underlying sources: the organic results, the People Also Ask questions, the knowledge graph. Give the agent those and it can ground its own answer and cite real URLs, which is what you wanted from "AI Mode" in the first place.
The Scavio Solution
Point your agent at Scavio's hosted MCP server at https://mcp.scavio.dev/mcp and authenticate with the x-api-key header. It runs in opencode, Claude Code, Cursor, VS Code, and Windsurf with a few lines of config. When your agent calls the Google tool, it gets back structured JSON from POST /api/v1/google: organic results, people_also_ask, knowledge_graph, and related_searches. No HTML parsing, no rotating proxies, no maintenance burden on you. Be clear on one thing: Scavio does not return Google's AI Overview or AI Mode text. That's deliberate, and it's the better deal. Instead of a summary you can't verify, you get the raw ranked sources the summary was built from. Your agent reads organic + people_also_ask + knowledge_graph and writes its own grounded answer with citations you can trust. That's more useful for a coding agent than a paraphrase it can't trace. Pricing is $0.005 per credit; a Google call costs 2 credits with light_request:false, or 1 when light. You get 50 free credits at signup, and $30/mo is 7,000 credits. Set it up once and it keeps working.
Before
Your agent depends on an unmaintained scraper MCP that returns garbage the day Google changes its markup.
After
Your agent calls a hosted MCP and gets structured Google JSON: organic, people_also_ask, knowledge_graph, related_searches.
Who It Is For
Developers who wanted a "Google AI Mode MCP" for their coding agent and need reliable, structured Google results to ground on instead.
Key Benefits
- Hosted MCP at https://mcp.scavio.dev/mcp, no scraper to maintain or proxy to rotate
- Structured Google SERP: organic, people_also_ask, knowledge_graph, related_searches
- Works in opencode, Claude Code, Cursor, VS Code, and Windsurf
- Real source URLs to ground and cite, not an unverifiable AI summary
- $0.005/credit, 2 credits per Google call, 50 free credits at signup
Python Example
# MCP config (opencode / Claude Code): point at the hosted server
# {
# "mcpServers": {
# "scavio": {
# "url": "https://mcp.scavio.dev/mcp",
# "headers": { "x-api-key": "YOUR_API_KEY" }
# }
# }
# }
# Or call the Google endpoint directly
import requests
resp = requests.post(
"https://api.scavio.dev/api/v1/google",
headers={"Authorization": "Bearer YOUR_API_KEY"},
json={"query": "best vector database for rag", "light_request": False},
)
data = resp.json()
print(data["organic"])
print(data["people_also_ask"])
print(data["knowledge_graph"])JavaScript Example
// MCP config (opencode / Claude Code): point at the hosted server
// {
// "mcpServers": {
// "scavio": {
// "url": "https://mcp.scavio.dev/mcp",
// "headers": { "x-api-key": "YOUR_API_KEY" }
// }
// }
// }
// Or call the Google endpoint directly
const resp = await fetch("https://api.scavio.dev/api/v1/google", {
method: "POST",
headers: {
"Authorization": "Bearer YOUR_API_KEY",
"Content-Type": "application/json",
},
body: JSON.stringify({ query: "best vector database for rag", light_request: false }),
});
const data = await resp.json();
console.log(data.organic);
console.log(data.people_also_ask);
console.log(data.knowledge_graph);