Skip to main content
DataDoe HubHub
(opens in a new tab)
ChatGPT (opens in a new tab)
Claude (opens in a new tab)

Using Hermes Agent with DataDoe MCP

Hermes Agent (opens in a new tab) is an open-source, self-hosted AI agent created by Nous Research (opens in a new tab). Unlike standard AI chat assistants (like ChatGPT) that start from scratch with every conversation, Hermes is an autonomous agent designed to remember its experiences and grow progressively smarter over time. It supports a desktop app, CLI, and TUI interfaces.

In this setup, DataDoe MCP acts as the data layer for Hermes. Your agent calls DataDoe tools to read live Amazon Seller, Vendor, and Ads data on demand — without manual exports or custom SP-API pipelines.

What you will build

By the end of this tutorial, you will have:

  • Hermes Agent installed and running on your machine with the desktop UI
  • An Anthropic LLM configured with your own API key
  • DataDoe MCP connected via the mcp-remote proxy in config.yaml
  • A verified first conversation querying your Amazon seller data
  • A concrete 7-day sales report for the example seller account, generated inside Hermes chat

Prerequisites

Step 1: Install Hermes

  1. Download the Hermes desktop installer for your operating system from hermes-agent.nousresearch.com (opens in a new tab).
  2. Run the installer and open the Hermes desktop app.
  3. Follow the initial onboarding prompts (you can skip the Nous Portal step — we will configure your own Anthropic key in the next step).

Step 2: Configure an Anthropic LLM

Hermes supports many LLM providers. This tutorial uses claude-sonnet-4-6 via a direct Anthropic API key, configured through the Hermes CLI.

In your terminal, run:

bash
1hermes model

This launches the interactive model picker. Follow the steps below.

Select provider: Anthropic

Use the arrow keys to navigate to Anthropic (Claude models via API key or Claude Code) and press ENTER to select it.

Choose authentication method: API key

Hermes asks how to authenticate with Anthropic. Select option 2 — Anthropic API key (pay-per-token):

Enter your Anthropic API key

Hermes prompts for the key and shows where to create one:

Get an API key at: https://platform.claude.com/settings/keys
API key (sk-ant-...): _

Paste your key from console.anthropic.com/settings/keys (opens in a new tab) and press ENTER.

Already have credentials stored? If you have previously authenticated, Hermes shows a Use existing credentials / Reauthenticate prompt instead. Choose Use existing credentials to skip re-entry.

Select default model: claude-sonnet-4-6

Pick claude-sonnet-4-6 from the model list (use / to search) and press ENTER. Hermes marks it as the default and writes the selection to ~/.hermes/config.yaml.

The resulting section in ~/.hermes/config.yaml:

yaml
1model:
2    provider: anthropic
3    default: claude-sonnet-4-6

Already inside a Hermes chat? Use /model claude-sonnet-4-6 --provider anthropic --global to hot-swap the running session and persist the change at the same time.

Step 3: Connect DataDoe MCP via mcp-remote

Hermes reads MCP server configuration from ~/.hermes/config.yaml. DataDoe MCP is a remote HTTP server, so you connect it through the mcp-remote proxy, which bridges the HTTP transport into the local stdio interface that Hermes expects.

  1. Open ~/.hermes/config.yaml in any text editor.
  2. Add or extend the mcp_servers section with the following block:
yaml
1mcp_servers:
2    datadoe:
3        command: 'npx'
4        args:
5            - '-y'
6            - 'mcp-remote'
7            - 'https://mcp.datadoe.com/mcp/v1'
8            - '--header'
9            - 'datadoe-mcp-key: YOUR_DATADOE_MCP_KEY'

Replace YOUR_DATADOE_MCP_KEY with your key from DataDoe MCP Integrations (opens in a new tab).

  1. Save the file and restart Hermes (quit and reopen the desktop app).

Quick setup reference

If you are already familiar with Hermes, here is everything you need:

Proxy:    mcp-remote  (npx -y mcp-remote ...)
URL:      https://mcp.datadoe.com/mcp/v1
Header:   datadoe-mcp-key: YOUR_DATADOE_MCP_KEY
Config:   ~/.hermes/config.yaml  →  mcp_servers.datadoe

Step 4: Verify the MCP connection

  1. Restart Hermes and open a new chat session.
  2. Type the following message:
What DataDoe tools do you have access to?

Hermes should respond with a list of DataDoe MCP tools such as list_sellers, query_table, and others. If no tools appear, check that Node.js is installed and that the mcp_servers block in config.yaml is correctly indented.

Step 5: First conversation with Amazon data

CLI or UI — your choice. You can chat with Hermes from the terminal (hermes chat) or from the Hermes desktop UI (opens in a new tab). Both interfaces share the same config.yaml and MCP setup. The examples from this step onwards are shown using the Hermes desktop UI client.

Run a simple test query to confirm that live Amazon data flows through DataDoe:

List all my connected sellers

Hermes calls the list_sellers DataDoe tool and returns the names of your connected accounts. If you see your sellers listed, the connection is working.

Step 6: Generate a 7-day sales report for Delto UK

Now use Hermes to produce a real business report. Open a new chat and paste the prompt below. Hermes will call DataDoe tools, retrieve the data, analyse it, and write a structured markdown report.

What are sales for last 7 days for Delto UK seller?

Hermes will:

  1. Call list_sellers to confirm the Delto UK account.
  2. Query the Profit by SKU & Date table for the last 7 days.
  3. Analyse the data and produce a structured markdown report with per-SKU breakdowns, daily rows, and executive KPIs.

Related resources

DataDoe MCP resources

Check the following resources for more information: