How to Set Up Hermes Agent with Tinyhat

Hermes Agent can live in the same place you already handle the rest of your day: Telegram. Tinyhat gives it a private, always-on cloud computer, connects it to your own Telegram bot, and handles the infrastructure behind it.

You do not need to install a server, keep a laptop awake, or spend an afternoon wiring services together. You start the computer, open Telegram, and talk to your agent.

The one-minute version

1. Open the Tinyhat page and start your Hermes setup.
2. Follow the guided flow to create your private Telegram bot and Tinyhat computer.
3. Open the bot in Telegram and answer the short introduction questions.
4. Start with the model credit included by Tinyhat.
5. If you want, connect a supported paid ChatGPT account through the secure authorization flow.
6. Connect only the tools Hermes needs for the work you want to delegate.
7. Give it one small, real task and build from there.

That is the whole setup. The important part comes next: deciding what Hermes should be allowed to do.

What Tinyhat sets up for you

Hermes is the agentic OS. Tinyhat provides the computer it runs on.

Each Tinyhat customer gets an isolated cloud computer rather than a shared agent workspace. Hermes runs there around the clock, even when your laptop is closed. Your private Telegram bot is the interface, so there is no separate dashboard you need to keep open just to ask for work.

The result feels simple in Telegram, but the separation matters. Your agent has a clean workspace of its own instead of broad access to your personal laptop.

Step 1: Start from the Tinyhat page

Open the Tinyhat page and choose Create your Hermes:

Start with Hermes on Tinyhat

The setup flow creates the Tinyhat computer and connects it to your private Telegram bot. Follow the prompts on screen until Telegram opens.

Step 2: Meet your agent in Telegram

Open the bot and complete the short introduction. Tell Hermes:

• What it should call itself
• What it should call you
• What kind of work you expect to delegate
• Any standing boundary that should apply from the beginning

You can change these details later. Do not spend twenty minutes trying to design the perfect personality. Give it enough context to be useful, then improve it through real work.

A good first boundary is:

You may research, organize, and draft freely. Ask before sending messages, publishing, deleting data, making purchases, or changing an external system.

Step 3: Choose how Hermes accesses AI models

Tinyhat includes managed model credit so you can start immediately. You do not need to bring an API key just to have the first conversation.

If you have a supported paid ChatGPT account, you can connect it later through Tinyhat's secure device-authorization flow. Hermes will guide you through the required ChatGPT security setting and give you the authorization step in Telegram. You never need to paste a password or raw access token into chat.

Model access and tool access are separate. Connecting ChatGPT gives Hermes access to supported subscription models. It does not automatically connect your email, calendar, Drive, or other services.

Step 4: Connect tools only when a task needs them

Hermes becomes useful when it can work with the systems you already use. Start from the workflow, then connect the minimum access required.

For example:

• Meeting preparation may need read access to Calendar and selected documents.
• Inbox cleanup may need Gmail read access before you allow labels, drafts, or sends.
• GitHub issue triage may need Triage access without code-write permission.
• Publishing may begin with drafts and require your approval before anything goes live.

Tinyhat supports native authorization flows for connected services. When another tool requires an API key or token, Hermes sends a secure entry button. The secret is entered outside the chat and stored on your private computer.

Do not connect every account on day one. Give Hermes one workflow, watch how it handles it, and expand access only when the benefit is clear.

Step 5: Give Hermes a real first task

A vague request like "help me with my business" gives the agent very little to work with. A bounded task gives you something concrete to evaluate.

Try one of these:

Review this document and give me the five decisions I need to make.

Prepare a brief for tomorrow's meetings, but do not send or change anything.

Turn these notes into a draft weekly update. Keep it under 300 words.

Inspect this GitHub issue backlog and recommend the first three issues to triage. Do not modify the repository yet.

The first result does not need to be perfect. React to it. Ask for a shorter version, a different structure, or better evidence. Hermes gets more useful when you work through a few real iterations instead of trying to write one perfect prompt.

Keep control as access grows

An agent that can use tools can affect real systems. Treat permissions as part of the workflow, not as a setup chore.

A practical progression is:

1. Let Hermes observe and summarize.
2. Let it recommend actions.
3. Let it prepare drafts for approval.
4. Automate only the low-risk steps that have already proven reliable.

This keeps the useful part of autonomy without asking you to trust everything at once.

You are ready

At this point Hermes has its own computer, a private Telegram interface, model access, and whatever tools you chose to connect. Start with one recurring task that wastes your time today. If the workflow works, teach Hermes how you want it repeated.

Create your Hermes on Tinyhat