Les agents Hermes prennent en charge l'appel d'outils via MCP mais sont livrés sans outil de recherche intégré. Ajouter Scavio MCP donne à votre agent Hermes accès aux données de recherche de Google, YouTube, Amazon, Walmart, Reddit et TikTok sans aucun code personnalisé. Le point de terminaison MCP est https://mcp.scavio.dev/mcp et coûte 0,005 $ par appel d'outil. Ce tutoriel le connecte en moins de 5 minutes.
Prérequis
- Runtime de l'agent Hermes installé
- Une clé API Scavio de scavio.dev
- Connaissance de base de la configuration Hermes
Parcours
Étape 1: Obtenez votre clé API Scavio et votre point de terminaison MCP
Inscrivez-vous sur scavio.dev et notez l'URL du serveur MCP.
Bash
# 1. Sign up at https://scavio.dev (free: 250 credits/month)
# 2. Dashboard > API Keys > Create New Key
# 3. MCP endpoint: https://mcp.scavio.dev/mcp
# Verify the MCP server is reachable:
curl -X POST https://mcp.scavio.dev/mcp \
-H 'x-api-key: YOUR_KEY_HERE' \
-H 'Content-Type: application/json' \
-d '{"jsonrpc": "2.0", "method": "tools/list", "id": 1}'Étape 2: Ajoutez Scavio MCP à la configuration de l'agent Hermes
Mettez à jour la configuration de votre agent Hermes pour inclure le serveur Scavio MCP.
# In your Hermes agent config (config.yaml or equivalent):
# Add the MCP server under the tools or mcp_servers section
mcp_servers:
scavio:
url: "https://mcp.scavio.dev/mcp"
headers:
x-api-key: "${SCAVIO_API_KEY}"
tools:
- search
- extract
- tiktok
# Environment variable:
# export SCAVIO_API_KEY=your_key_hereÉtape 3: Testez l'outil de recherche dans votre agent
Exécutez une requête de test pour vérifier que la connexion MCP fonctionne de bout en bout.
Python
# Start your Hermes agent with the updated config
# Then send a test message:
# User: Search for the best Python web frameworks in 2026
# Expected agent behavior:
# 1. Agent recognizes it needs live data
# 2. Calls scavio.search tool via MCP
# 3. Receives structured Google results
# 4. Synthesizes answer from live data
# You can also test programmatically:
import requests
SH = {'x-api-key': os.environ['SCAVIO_API_KEY'], 'Content-Type': 'application/json'}
data = requests.post('https://api.scavio.dev/api/v1/search',
headers=SH, json={'query': 'best python web framework 2026', 'country_code': 'us'}).json()
print(f'Results: {len(data.get("organic_results", []))} organic')
print(f'AI Overview: {"yes" if data.get("ai_overview") else "no"}')Étape 4: Configurez les recherches spécifiques à une plateforme
Utilisez des paramètres de plateforme pour rechercher sur Reddit, YouTube, Amazon, etc.
Bash
# Your Hermes agent can now use these search patterns:
# Google search (default):
# {"query": "serp api comparison", "country_code": "us"}
# Reddit search:
# {"query": "best api for agents", "platform": "reddit", "country_code": "us"}
# YouTube search:
# {"query": "python tutorial 2026", "platform": "youtube", "country_code": "us"}
# Amazon product search:
# {"query": "mechanical keyboard", "platform": "amazon", "country_code": "us"}
# Example agent prompt that triggers multi-platform search:
# "Research what Reddit says about the best SERP APIs,
# then check YouTube for tutorial coverage"
# Cost: $0.005 per search call
# Free tier: 250 searches/month
# Project plan: $30/month = 7,000 searchesExemple Python
Python
# The MCP integration requires no Python code.
# To verify the same data programmatically:
import os, requests
SH = {'x-api-key': os.environ['SCAVIO_API_KEY'], 'Content-Type': 'application/json'}
for platform in ['google', 'reddit', 'youtube']:
data = requests.post('https://api.scavio.dev/api/v1/search',
headers=SH, json={'query': 'best serp api', 'platform': platform if platform != 'google' else None,
'country_code': 'us'}).json()
print(f'{platform}: {len(data.get("organic_results", []))} results ($0.005)')Exemple JavaScript
JavaScript
// MCP integration requires no JS code.
// To verify the same data programmatically:
const SH = { 'x-api-key': process.env.SCAVIO_API_KEY, 'Content-Type': 'application/json' };
for (const platform of [null, 'reddit', 'youtube']) {
const body = { query: 'best serp api', country_code: 'us' };
if (platform) body.platform = platform;
const data = await fetch('https://api.scavio.dev/api/v1/search', {
method: 'POST', headers: SH, body: JSON.stringify(body)
}).then(r => r.json());
console.log(`${platform || 'google'}: ${(data.organic_results || []).length} results`);
}Sortie attendue
JSON
# After config update and agent restart:
Hermes Agent > MCP Servers:
scavio: Connected (3 tools available)
# Test query:
User: Search for best Python web frameworks in 2026
Agent: [Calling scavio.search]
Based on current search results:
1. FastAPI leads for async API development
2. Django 5.1 remains the full-stack standard
3. Litestar gaining traction as FastAPI alternative
google: 10 results ($0.005)
reddit: 8 results ($0.005)
youtube: 10 results ($0.005)