Hermes Agent Web Search Setup Guide

Hermes Agent ships without built-in web search. Adding live search capability requires connecting an external search provider via MCP or a custom tool. The three practical options in 2026: Tavily MCP (1K free searches/mo, AI-summarized results), Scavio MCP (250 free credits/mo, structured multi-platform results), or self-hosted SearXNG (free, unlimited, requires a server).

Option 1: Scavio MCP (structured multi-platform)

Returns structured JSON from Google, Amazon, YouTube, Reddit, Walmart, and TikTok. Best for agents that need to parse specific fields (prices, ratings, URLs) rather than read summaries.

# .mcp.json configuration for Scavio
{
  "mcpServers": {
    "scavio": {
      "url": "https://mcp.scavio.dev/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_SCAVIO_API_KEY"
      }
    }
  }
}

# Or use the HTTP API directly as a custom tool
import requests, os

def web_search(query: str, platform: str = "google") -> dict:
    """Search tool for Hermes Agent."""
    resp = requests.post("https://api.scavio.dev/api/v1/search",
        headers={"x-api-key": os.environ["SCAVIO_API_KEY"]},
        json={"query": query, "platform": platform})
    return resp.json()

Option 2: Tavily MCP (AI-summarized)

Returns AI-generated summaries of search results. Best for agents that need context for reasoning, not raw data for extraction. 1K free searches/mo, then $0.008/credit (basic search = 1 credit, advanced = 2).

# Tavily as a custom tool
from tavily import TavilyClient

tavily = TavilyClient(api_key=os.environ["TAVILY_API_KEY"])

def tavily_search(query: str) -> dict:
    """AI-summarized search for Hermes Agent."""
    return tavily.search(query=query, search_depth="basic")

# Returns: {"answer": "summarized answer", "results": [...]}

Option 3: SearXNG (self-hosted, free)

Open-source meta-search engine. Free and unlimited, but requires hosting. Results are less structured than commercial APIs. Best for budget-constrained setups where you control the infrastructure.

# Deploy SearXNG with Docker
docker run -d --name searxng -p 8080:8080 searxng/searxng:latest

# Query it from your agent
curl "http://localhost:8080/search?q=best+crm+2026&format=json"

Which to pick

• Agent extracts structured data (prices, ratings, URLs): Scavio -- structured JSON, multi-platform
• Agent needs context for reasoning: Tavily -- AI summaries reduce token usage
• Agent runs high volume on a budget: SearXNG -- free but requires server maintenance
• Agent needs multiple platforms (Amazon + Reddit + Google): Scavio -- single API, multiple platforms

Cost comparison at 500 searches/mo

• Scavio: 250 free + 250 x $0.005 = $1.25/mo
• Tavily: 500 within free tier (1K/mo) = $0/mo
• SearXNG: $5-10/mo server cost, unlimited queries
• At 2,000 searches/mo: Scavio $8.75, Tavily $8 (1K free + 1K x $0.008), SearXNG still $5-10