Hermes Agent is Nous Research's 2026 autonomous task-execution framework. It ships with 40+ built-in tools and native MCP support, which means you can extend it with Scavio for live web, shopping, video, and discussion search. This tutorial walks through wiring Scavio's MCP server into a Hermes setup running on local Qwen 3.5.
Prerequisites
- Hermes Agent installed from hermes-agent.nousresearch.com
- A local Qwen 3.5+ instance via Ollama or vLLM
- A Scavio API key
- Node.js 20+ for the MCP server
Walkthrough
Step 1: Install Hermes Agent
Download and install Hermes Agent per the official docs.
Bash
# Follow instructions at hermes-agent.nousresearch.comStep 2: Start your local LLM
Launch Qwen 3.5 via Ollama.
Bash
ollama run qwen3.5:4bStep 3: Register Scavio in the Hermes tool gateway
Edit ~/.hermes/tools.json to add the Scavio MCP server.
JSON
{
"mcpServers": {
"scavio": {
"command": "npx",
"args": ["-y", "@scavio/mcp"],
"env": { "SCAVIO_API_KEY": "sk_live_..." }
}
}
}Step 4: Launch Hermes with the tool gateway
Start Hermes pointing at your local Qwen endpoint.
Bash
hermes --model http://localhost:11434 --tools scavioStep 5: Run an autonomous task
Ask Hermes to do research that requires live data.
Bash
hermes task 'Find the top 5 Reddit posts about LangGraph this week and summarize the common complaints'Python Example
Python
# Hermes runs autonomously. To validate Scavio independently:
import os, requests
r = requests.post('https://api.scavio.dev/api/v1/search',
headers={'x-api-key': os.environ['SCAVIO_API_KEY']},
json={'platform': 'reddit', 'query': 'langgraph', 'sort': 'new', 'time': 'week'})
print(r.json()['posts'][:5])JavaScript Example
JavaScript
const res = await fetch('https://api.scavio.dev/api/v1/search', {
method: 'POST',
headers: { 'x-api-key': process.env.SCAVIO_API_KEY, 'Content-Type': 'application/json' },
body: JSON.stringify({ platform: 'reddit', query: 'langgraph', sort: 'new', time: 'week' })
});
console.log((await res.json()).posts.slice(0, 5));Expected Output
JSON
Hermes runs through a multi-step plan: scavio.search_reddit for LangGraph posts, scavio.get_reddit_post for top comments on each, then synthesizes the common complaints into a report.