native-mcp

Discover and register MCP server tools as native Hermes Agent tools.

1|Updated Apr 14, 2026
One-click install
npx skills add https://github.com/eddielueng/hermes-agent-zh --skill native-mcp-eddielueng
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: native-mcp
Source: https://github.com/eddielueng/hermes-agent-zh/tree/main/skills/mcp/native-mcp
Command: npx skills add https://github.com/eddielueng/hermes-agent-zh --skill native-mcp-eddielueng

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

It eliminates the friction of configuring and bridging external capabilities into Hermes Agent by automatically discovering MCP server tools and making them available as native agent tools.

Core Features & Use Cases

  • MCP tool auto-discovery and registration: Connects to configured MCP servers at startup, lists available tools, and registers them into Hermes’s tool registry.
  • Native tool injection across platforms: Injects discovered MCP tools into all Hermes platform toolsets (so they’re usable in any conversation).
  • Flexible transports and resilient connections: Supports both stdio-based MCP servers (spawned subprocess) and HTTP/StreamableHTTP MCP servers, including automatic reconnection and configurable timeouts.
  • Security guardrails: Filters environment variables for stdio subprocesses and redacts credential-like patterns from failing tool errors to reduce leakage risk.

Quick Start

Configure your MCP server in ~/.hermes/config.yaml under mcp_servers, then restart Hermes Agent to automatically expose discovered tools as native commands the agent can call.

Frequently Asked Questions about native-mcp

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I connect MCP servers to an agent and use their tools natively?

You can connect MCP servers to an agent by configuring them in ~/.hermes/config.yaml under mcp_servers, which enables automatic discovery and registration of their tools as native callable agent commands at startup.

Can I use HTTP transport for MCP server integration, or is it limited to stdio?

MCP server integration supports both stdio-based servers via spawned subprocesses and HTTP/StreamableHTTP transports, with the latter requiring MCP SDK support for resilient connections and configurable timeouts.

How does MCP tool discovery work when connecting external servers?

MCP tool discovery works by connecting to configured MCP servers at startup, listing their available tools, and injecting them into the agent's tool registry so they are usable across all platform conversations.

What security guardrails apply when integrating MCP servers with subprocesses?

Security guardrails for MCP server integration include filtering environment variables for stdio subprocesses and automatically redacting credential-like patterns from failing tool errors to reduce sensitive data leakage risk.

Do I need a separate bridge workflow to expose MCP tools across different agent platforms?

You do not need a separate bridge workflow; discovered MCP tools are natively injected into all Hermes platform toolsets, providing persistent tool availability across all environments without additional bridging configuration.

How are MCP server tools named when registered as native agent tools?

MCP server tools are registered using a deterministic naming convention of mcp_{server_name}_{tool_name}, ensuring predictable and consistent tool identification across the agent's tool registry.