native-mcp

Connect Zeus Agent to external MCP servers and register their tools.

Updated May 5, 2026
One-click install
npx skills add https://github.com/Z43L/zeus-agent --skill native-mcp-z43l
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: native-mcp
Source: https://github.com/Z43L/zeus-agent/tree/main/skills/mcp/native-mcp
Command: npx skills add https://github.com/Z43L/zeus-agent --skill native-mcp-z43l

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires mcp.

What problem does it solve?

It eliminates the manual work of bridging external MCP servers into Zeus Agent by automatically connecting, discovering, and registering MCP-provided tools for direct use in every conversation.

Core Features & Use Cases

  • Automatic MCP server discovery at startup: Reads your configured MCP servers and loads their available tools into the agent’s tool registry.
  • First-class tool availability with naming: Exposes tools with the mcp_{server_name}_{tool_name} prefix (sanitizing names for compatibility).
  • Works with both stdio and HTTP/StreamableHTTP transports: Runs local stdio-based servers (via npx/uvx) and connects to remote HTTP-based MCP servers.
  • Tool-call robustness and safety: Uses persistent background connections, retries with exponential backoff, filters stdio subprocess environment variables, and redacts credential-like strings in errors.
  • Optional sampling support: Enables MCP server-initiated LLM “sampling” requests through configurable per-server settings, including tool-round limits to prevent infinite loops.

Quick Start

Configure your MCP servers under the mcp_servers key in ~/.hermes/config.yaml, then restart Zeus Agent so it discovers tools and registers them as mcp_{server_name}_{tool_name}.

Frequently Asked Questions about native-mcp

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

FAQPage Schema
How do I connect external MCP servers to my agent for tool discovery?

To connect external MCP servers, you configure them under the `mcp_servers` key in `~/.hermes/config.yaml`, and the agent automatically performs startup tool discovery, registering their tools for direct conversation use.

Can I use both stdio subprocesses and HTTP endpoints for MCP server integration?

MCP server integration supports both local stdio-based subprocesses launched via npx/uvx and remote HTTP/StreamableHTTP endpoints, handling persistent async connections and retries automatically across both transport types.

How does MCP tool naming work when registered inside the agent?

Discovered MCP tools are registered using an `mcp_{server_name}_{tool_name}` naming convention, sanitizing the original tool names for compatibility so you can call them directly within your agent conversations.

Does MCP sampling support allow servers to initiate LLM requests safely?

MCP sampling allows servers to initiate LLM sampling requests through configurable per-server settings, applying tool-round limits to prevent infinite loops while maintaining controlled message flows.

How are credentials protected when stdio subprocess errors occur during MCP connections?

Credential safety during stdio subprocess errors is handled by filtering environment variables and redacting credential-like strings in error messages, ensuring sensitive data is not exposed in connection failures.

What is the best way to integrate filesystem and database capabilities via MCP?

Integrating filesystem, GitHub, database, and API capabilities via MCP requires configuring the respective servers in your config file, letting the agent connect and register their tools automatically for immediate use.