Hermes MCP Catalog: One-Click Agent Tool Integrations

Key Takeaways

• Run hermes mcp to open an interactive picker showing available, installed, and enabled MCP tools — no more manual config files.
• Every tool in the catalog is pre-reviewed by the Nous Research team before it can be installed; there is no open submission queue.
• Allowlist or blocklist specific tool actions: if a server exposes 20 actions but you only want 3, you can restrict it to exactly those.
• Auto-update feature: when a connected tool gains new capabilities at runtime, Hermes detects and picks them up without requiring a restart.
• Hermes works bidirectionally — it can use MCP tools as a client, and it can also expose itself as an MCP server so other agents can tap into its connected channels.
• Parallel tool execution is available for read-only tools that won't conflict; sequential execution remains the default for write operations.

What MCP Is and Why This Update Matters

MCP (Model Context Protocol) is the standard way to connect external tools — GitHub, databases, file systems, workflow apps — to an AI agent. Previously, wiring a tool into Hermes required digging through documentation, copying config files, and hoping nothing broke. That friction stopped most non-technical users from getting past the first tool.

The new catalog removes that barrier entirely. You run hermes mcp, an interactive menu appears listing every approved integration, and you install with a single keypress. Hermes handles the setup, asks for any required API keys, and stores them securely. The tool is active immediately.

Security Model

The catalog's biggest practical benefit is the security review gate. Unlike connecting any arbitrary MCP server from the internet, catalog tools have been inspected by the Nous team. Combined with the tool-level allowlist feature — where you can whitelist only the specific actions you actually want your agent to use — this gives you defense in depth. Your agent cannot touch capabilities you haven't explicitly enabled.

The sampling feature (where a connected tool can borrow Hermes's model for reasoning) has built-in usage limits, so a rogue tool cannot exhaust your agent's resources.

Commands & Code Mentioned

hermes mcp

Related on OpenClawDatabase

• Hermes Agent Hub — overview, setup, and all guides
• Hermes MCP Tools Guide — connecting and managing MCP integrations
• Security Center — cross-platform agent security guidance
• Compare agent platforms — Hermes vs OpenClaw vs Claude Cowork