How is this workflow triggered?

This workflow uses a automatic on cursor agent search tool call. Automatic on agent search calls.

Which Scavio platforms does this workflow use?

This workflow uses the following Scavio platforms: google. Each platform is called via the same unified API endpoint.

Can I run this workflow on the free tier?

Yes. Scavio's free tier includes 250 credits per month with no credit card required. That is enough to test and validate this workflow before scaling it.

Cursor Agent Search Grounding Workflow

Overview

Cursor's background agent generates code from stale training data. This workflow adds MCP-based web search so the agent can verify package versions, check current API documentation, and confirm library function signatures before writing code. Reduces the rate of deprecated API suggestions and wrong version references.

Trigger

Automatic on Cursor agent search tool call

Schedule

Automatic on agent search calls

Workflow Steps

Configure MCP server

Add the search MCP server to .cursor/mcp.json with your API key. Cursor automatically discovers and exposes the search tool to the agent.

Agent detects knowledge gap

During code generation, the agent recognizes when it needs current information: package versions, API endpoints, or library signatures.

Search for current docs

The agent calls the MCP search tool with a targeted query like 'react-router v7 useNavigate API' to get current documentation.

Verify and apply

The agent cross-references search results with its training data, preferring current documentation when they conflict.

Python Implementation

Python

# .cursor/mcp.json configuration
# {
#   "mcpServers": {
#     "search": {
#       "url": "https://mcp.scavio.dev/mcp",
#       "headers": { "x-api-key": "your-key" }
#     }
#   }
# }

# The agent uses the MCP tool automatically. Here's how the search works internally:
import requests, os

H = {"x-api-key": os.environ["SCAVIO_API_KEY"], "Content-Type": "application/json"}

def verify_package_version(package_name, language="python"):
    """Check current package version before installing."""
    r = requests.post("https://api.scavio.dev/api/v1/search", headers=H,
        json={"platform": "google",
              "query": f"{package_name} latest version {language} 2026"}).json()
    for o in r.get("organic", [])[:3]:
        snippet = o.get("snippet", "")
        if "version" in snippet.lower() or "v" in snippet.lower():
            return {"package": package_name, "source": o.get("title", ""),
                    "snippet": snippet[:150]}
    return {"package": package_name, "source": "not found", "snippet": ""}

info = verify_package_version("langchain")
print(f"{info['package']}: {info['snippet']}")

JavaScript Implementation

JavaScript

// .cursor/mcp.json
// {
//   "mcpServers": {
//     "search": {
//       "url": "https://mcp.scavio.dev/mcp",
//       "headers": { "x-api-key": "your-key" }
//     }
//   }
// }

const H = {"x-api-key": process.env.SCAVIO_API_KEY, "Content-Type": "application/json"};

async function verifyPackageVersion(packageName, language = "javascript") {
  const r = await fetch("https://api.scavio.dev/api/v1/search", {
    method: "POST", headers: H,
    body: JSON.stringify({platform: "google",
      query: `${packageName} latest version ${language} 2026`})
  }).then(r => r.json());
  for (const o of (r.organic || []).slice(0, 3)) {
    const snippet = o.snippet || "";
    if (snippet.toLowerCase().includes("version") || snippet.includes("v")) {
      return {package: packageName, source: o.title || "", snippet: snippet.slice(0, 150)};
    }
  }
  return {package: packageName, source: "not found", snippet: ""};
}

verifyPackageVersion("next").then(i =>
  console.log(`${i.package}: ${i.snippet}`)
);

Platforms Used

Google

Web search with knowledge graph, PAA, and AI overviews

Cursor Agent Search Grounding Pipeline