Hermes Agent Connectors: Chat Apps, Models, and Tools

Hermes Agent is only as useful as what it connects to. It links to three different kinds of thing, and people often mix them up.

The first is message gateways, the chat apps it talks to you through. The second is model providers, the AI brains that power it. The third is tools and data, the systems it acts on.

This guide covers all three, with the setup path for each. It also explains how the Model Context Protocol lets you add connections that are not built in.

What can Hermes Agent connect to?

Hermes Agent connects to message gateways, model providers, and external tools, and each type is set up differently. Knowing which is which saves a lot of confusion.

Message gateways are how the agent reaches you, such as Telegram or Slack. Model providers are the AI models that give it intelligence, such as OpenAI or a local model.

Tools and data are the systems it works with through the Model Context Protocol and its own skills. Together these three connector types decide what the agent can see and do.

• Message gateways — the chat apps it replies on.
• Model providers — the AI models that power its thinking.
• Tools and data — the systems it reads from and acts on.

How do you connect chat apps (message gateways)?

You connect chat apps with the hermes gateway setup command, then start them with hermes gateway start. A single gateway process handles all of them at once.

This is what lets you message your agent from your phone and get replies wherever you already chat. It turns any supported app into a front door for the agent.

Hermes Agent supports the major messaging platforms plus email and home automation. You can enable one or several, depending on where your team already works.

• Telegram — the most common choice for quick mobile access.
• Slack and Discord — good for team channels.
• WhatsApp and Signal — for private, direct messaging.
• Email — for reports and summaries that belong in your inbox.
• Home Assistant — referenced in the docs for home-automation triggers.

How do you connect model providers, including local models?

You connect a model provider with the hermes model command, which lets you choose a provider and add credentials. Hermes Agent is model-neutral, so you are never locked in.

For a managed start, run hermes setup --portal to log into Nous Portal over OAuth and reach many models through one login. You can switch to another provider anytime.

This is the connector that matters most for privacy. Pointing Hermes Agent at a local model through Ollama or LM Studio keeps your data on your own hardware, which is the safest path for regulated or confidential work.

• Nous Portal — one login for many models.
• OpenAI and OpenRouter — hosted frontier and multi-model access by API key.
• Local via Ollama or LM Studio — private, on-device models with no data leaving your machine.
• Custom endpoint — any OpenAI-compatible API you already run.

Why connect Hermes Agent to a local model?

You connect Hermes Agent to a local model when data privacy or compliance rules out sending information to a hosted API. A local model keeps everything on hardware you control.

For a business in a regulated field, this is often the difference between being able to use an autonomous agent and not. The agent, its memory, and your data all stay in-house.

The trade-off is capability and hardware cost. A local model may be less powerful than a hosted frontier model, and it needs a machine with enough memory. We help teams weigh that balance in an AI governance review.

How do you connect tools and data with MCP?

You connect external tools and data to Hermes Agent through the Model Context Protocol, an open standard for giving agents access to systems. The docs describe connecting any MCP server for extended capabilities.

The Model Context Protocol is how you safely extend what the agent can reach beyond its built-in abilities. If a system has an MCP server, Hermes Agent can use it.

The project ships optional MCP connectors you can turn on, and the agent's own skills add reusable procedures on top. Together they let the agent read files, search the web, and work with the systems you connect.

• Model Context Protocol — the open standard for connecting any MCP server.
• Optional MCPs — the project bundles connectors you can enable as needed.
• Built-in tools — file handling, web search, and command running come standard.
• Skills — reusable procedures the agent builds for jobs it repeats.

Which connectors should you set up first?

Set up one model provider and one chat app first, then add tools once the basics work. Trying to wire everything at once is the most common early mistake.

Start with a model, because the agent cannot think without one. Add a single chat app like Telegram so you can reach it from your phone.

Only then add tool and data connectors through the Model Context Protocol, and grant them carefully. Every new connector widens what the agent can touch, so treat access as something you add on purpose, not by default.

• Step 1: connect a model with hermes model.
• Step 2: connect one chat app with hermes gateway setup.
• Step 3: add tools through MCP only once the core loop works.
• Rule: grant the least access the job actually needs.