MCP Configuration

Connect Model Context Protocol servers for live tool access — databases, APIs, browsers, and more.

What is MCP?

MCP (Model Context Protocol) is an open standard developed by Anthropic that allows AI agents to connect to external tools, data sources, and services. Instead of each agent framework building its own tool integration system, MCP provides a universal protocol that any MCP-compatible agent can use.

Each MCP server is a lightweight process that exposes tools, resources, or prompts to the AI agent. Your Hermes agent discovers and calls these tools automatically — no manual plugin code needed.

Quick Start

Step 1: Add an MCP server to your Hermes configuration:

# File system access
hermes mcp add filesystem \
  --command "npx -y @modelcontextprotocol/server-filesystem /home/YOU/agentlife"

Step 2: Test the connection:

hermes mcp test filesystem

Step 3: Restart your agent session:

/reset

Your agent can now read and write files in the specified directory.

Common MCP Servers

Server	Purpose	Auth	Install
filesystem	Read/write files and directories	🔓 None	npx -y @modelcontextprotocol/server-filesystem <path>
web_search	Web search and content extraction	🔑 API key	npx -y tavily-mcp
sqlite	Query and manage SQLite databases	🔓 None	uvx mcp-server-sqlite --db <path>
github	GitHub repos, issues, PRs, code	🔑 Token	npx -y @modelcontextprotocol/server-github
puppeteer	Headless browser automation	🔓 None	npx -y @modelcontextprotocol/server-puppeteer
gmail	Gmail read, send, search	🔑 OAuth	npx -y @modelcontextprotocol/server-gmail
notion	Notion pages, databases, search	🔑 Token	npx -y @modelcontextprotocol/server-notion
google_docs	Google Docs create and edit	🔑 OAuth	npx -y @modelcontextprotocol/server-google-docs
arxiv	Academic paper search	🔓 None	uvx mcp-server-arxiv
youtube_transcript	YouTube video transcripts	🔓 None	uvx mcp-server-youtube-transcript
mermaid	Diagrams and flowcharts	🔓 None	uvx mermaid-mcp-server
stripe	Payment processing, revenue	🔑 Key	npx -y @modelcontextprotocol/server-stripe
time	Timezone and date utilities	🔓 None	uvx mcp-server-time

Per-Persona Recommendations

Each persona pack recommends specific MCP servers. Start with the ones marked 🔓 (no API key needed) and add 🔑 servers as you go.

Persona	Recommended MCP Servers
Life Ops	plaid, actual_budget, google_calendar, sqlite, puppeteer, gmail, web_search
Enterprise Architect	filesystem, web_search, github, mermaid, aws_pricing, azure_pricing, well_architected, confluence, jira
Knowledge Worker	arxiv, brave_search, google_docs, notion, confluence, sqlite, filesystem
Hobbyist	github, puppeteer, youtube_transcript, image_gen, mermaid, filesystem, web_search
Small Biz Owner	sqlite, gmail, puppeteer, google_docs, notion, filesystem, web_search
Student	arxiv, youtube_transcript, notion, google_docs, web_search, sqlite, filesystem
Indie Hacker	github, stripe, puppeteer, notion, web_search, sqlite, filesystem

Configuration File

MCP servers can also be configured directly in your Hermes config.yaml:

mcp_servers:
  filesystem:
    command: npx
    args:
      - -y
      - "@modelcontextprotocol/server-filesystem"
      - /home/YOU/agentlife
  web_search:
    command: npx
    args:
      - -y
      - tavily-mcp
    env:
      TAVILY_API_KEY: "${TAVILY_API_KEY}"

Fallback Modes

Every persona use case is designed to work without MCP servers. The agent falls back to manual input, file-based data, or web research. This means you can start using any persona immediately and add MCP servers later as your comfort level grows.

Common fallback modes:

• local_markdown — agent works with markdown files you provide
• manual_sources — you provide URLs and documents
• csv_import — agent reads data from CSV files you drop in a directory
• manual_tracking — you provide numbers and updates when prompted