The Problem
Multi-agent systems with critic/verifier loops need a data layer that the critic agent can query to fact-check claims made by the generator agent. Without live data access, the critic can only check for internal consistency, not factual accuracy. Most multi-agent frameworks (LangGraph, CrewAI, AutoGen) assume you will add tools, but do not ship with a reliable search tool that returns structured data suitable for verification.
The Scavio Solution
Add Scavio as the critic agent's verification tool in your multi-agent framework. When the generator produces a claim (e.g., 'SerpAPI costs $50/month'), the critic queries Scavio for the current data and compares. The structured JSON response makes verification deterministic: the critic checks specific fields rather than interpreting prose. This creates a grounded critic loop where factual claims are verified against live web data.
Before
Before the data layer, the critic agent could only check if the generator's claims were internally consistent. It could not verify pricing, dates, feature lists, or any factual claim against reality. The critic caught logical errors but passed through confident hallucinations.
After
After adding Scavio as the critic's tool, the system catches factual errors in real-time. Pricing claims are verified against live search results. The generator's output quality improved because the critic rejects hallucinated facts, forcing regeneration with correct data.
Who It Is For
AI engineers building multi-agent systems with critic/verifier patterns. Teams using LangGraph, CrewAI, or AutoGen who need their critic agent to verify factual claims against live data.
Key Benefits
- Critic agent verifies factual claims against live web data
- Structured JSON enables deterministic verification logic
- Works with LangGraph, CrewAI, AutoGen, and custom frameworks
- Per-query pricing means critic loops cost pennies per verification
- Catches confident hallucinations that internal consistency checks miss
Python Example
import requests
API_KEY = "your_scavio_api_key"
def verify_claim(claim: str, search_query: str) -> dict:
"""Critic agent tool: verify a claim against live web data."""
res = requests.post(
"https://api.scavio.dev/api/v1/search",
headers={"x-api-key": API_KEY},
json={"platform": "google", "query": search_query, "ai_overview": True},
timeout=15,
)
res.raise_for_status()
data = res.json()
evidence = []
if data.get("ai_overview"):
evidence.append({"source": "ai_overview", "text": data["ai_overview"]["text"][:500]})
for r in data.get("organic", [])[:3]:
evidence.append({"source": r.get("link", ""), "text": r.get("snippet", "")})
return {
"claim": claim,
"query": search_query,
"evidence_count": len(evidence),
"evidence": evidence,
}
# Critic agent verifies generator claim
result = verify_claim(
claim="SerpAPI costs $50/month for 5,000 searches",
search_query="SerpAPI pricing 2026"
)
print(f"Found {result["evidence_count"]} evidence sources for claim verification")
for e in result["evidence"]:
print(f" [{e["source"]}]: {e["text"][:100]}")JavaScript Example
const API_KEY = "your_scavio_api_key";
async function verifyClaim(claim, searchQuery) {
const res = await fetch("https://api.scavio.dev/api/v1/search", {
method: "POST",
headers: { "x-api-key": API_KEY, "content-type": "application/json" },
body: JSON.stringify({ platform: "google", query: searchQuery, ai_overview: true }),
});
const data = await res.json();
const evidence = [];
if (data.ai_overview) evidence.push({ source: "ai_overview", text: data.ai_overview.text.slice(0, 500) });
for (const r of (data.organic ?? []).slice(0, 3)) evidence.push({ source: r.link ?? "", text: r.snippet ?? "" });
return { claim, evidence };
}
const result = await verifyClaim("SerpAPI costs $50/month", "SerpAPI pricing 2026");
console.log(`Evidence sources: ${result.evidence.length}`);
result.evidence.forEach((e) => console.log(` [${e.source}]: ${e.text.slice(0, 100)}`));Platforms Used
Web search with knowledge graph, PAA, and AI overviews