Remote OpenClaw Blog
How to Integrate Third-Party Plugins with Hermes Agent
5 min read ·
You integrate third-party plugins with Hermes Agent through MCP (Model Context Protocol). Hermes connects to external tool servers — GitHub, databases, file systems, browsers, and internal APIs — and MCP support ships built-in with the standard install, so in most setups there is nothing extra to install before you add your first plugin.
How Plugins Work in Hermes
In Hermes Agent, a third-party plugin is an MCP server the agent connects to and auto-registers tools from. Rather than writing a native Hermes tool for every integration, you point Hermes at an existing MCP server and it discovers the tools that server exposes.
According to the official MCP documentation, Hermes performs automatic tool discovery and registration at startup, so once a server is configured its tools become available to the agent without further wiring. This is the same protocol used across Claude Code, Cursor, and Codex, which means most MCP servers work with Hermes unchanged. For a primer on the protocol itself, see our guide to the best MCP servers.
Stdio vs HTTP Plugin Servers
Hermes accepts two kinds of MCP servers in the same configuration: local stdio servers launched as a command, and remote HTTP servers reached by URL. The right choice depends on where the tool runs.
| Type | How Hermes reaches it | Best for |
|---|---|---|
| stdio (local) | Runs a local command (e.g. an npx package) and talks over stdin/stdout | File system, local databases, tools that need machine access |
| HTTP (remote) | Connects to a hosted server via a URL, with optional auth | Hosted APIs, shared team tools, SaaS integrations |
The MCP config reference documents both in a single file, so you can mix a local file-system server and a remote database server in one agent.
How to Add a Third-Party Plugin
You add a plugin with Hermes' add command, which registers the server and runs automatic tool discovery. The command accepts --url for HTTP servers, --command for stdio servers, --auth for credentials, and --args to pass the remaining arguments to a stdio command.
A remote HTTP server is added by URL; a local stdio server is added by the command that launches it. If you installed Hermes without extras, add MCP support first with the mcp extra, then register your server. Full command syntax lives in the Hermes CLI reference. Because tools register at startup, restart the agent after adding a server so the new tools load into the session — a step many people miss when a freshly added plugin does not appear.
One-Click Installs from the Catalog
Hermes ships a Nous-approved catalog that offers one-click installs and prompts inline for any keys a server needs. For popular integrations, this is faster than assembling a config by hand: you pick the server, Hermes installs it, and it asks for the API key at install time rather than failing silently later.
The catalog is a good default for common tools, while the add command remains the path for any MCP server not in the catalog — including your own internal servers. Either way, the result is the same: a registered server whose tools the agent can call. To find servers worth adding, our MCP directory lists thousands with install commands and source links.
Enabling, Disabling, and Managing Plugins
Hermes lets you enable or disable individual plugins without removing them, through an interactive toggle or scriptable flags. The plugins documentation notes options like --enable and --no-enable for scripted installs, so you can keep a server configured but inactive until you need it.
This matters for context and cost: every enabled server's tool descriptions are available to the agent, so disabling servers you are not using keeps the tool list focused. If you are also managing agent skills alongside plugins, see our Hermes Agent skills guide for how skills and tools interact.
Limitations and Tradeoffs
MCP integration is powerful but not free of tradeoffs. Each enabled server adds its tool descriptions to what the agent considers, so a large number of active plugins can crowd the context and slow tool selection — enable only what a given agent actually needs. Remote HTTP servers introduce a network dependency: if the server is down or rate-limited, those tools fail for the session. And third-party servers run with whatever access you grant them, so review a server's source before connecting it to sensitive systems, exactly as you would with any MCP server. When you only need one or two integrations, a single well-chosen server beats wiring up ten.
Related Guides
- Hermes Agent Skills: The Complete Guide
- Best Places to Find Hermes Agent Skills
- How to Set Up Hermes Agent from Scratch
- Awesome MCP Servers: The Best of 2026
Go deeper
The operator playbooks
Production-ready PDF guides for OpenClaw and Hermes Agent — $19.99 each.
Skills for this topic
Browse all skills →Frequently Asked Questions
How do I integrate third-party plugins with Hermes?
You connect them as MCP servers. Use Hermes' add command with --url for a remote HTTP server or --command for a local stdio server; Hermes discovers and registers the tools automatically at startup. Restart the agent so the new tools load.
Why doesn't my newly added plugin show up?
Tools register at startup, so a plugin added mid-session usually will not appear until you restart the agent. If it still does not appear after a restart, confirm the server is enabled and that its command or URL is correct.





