Opening Your Agent

This page used to describe a single control panel inside the OneClickClaw dashboard. That panel is gone. Today each job is done in one of two places:

• Inside your agent: the AI model, channels and conversations. Click Open Agent on your server card to get there. On OpenClaw this opens the OpenClaw chat interface; on Hermes Agent it opens your Hermes dashboard.
• On the server page: the cards that point into your agent, the personality files (OpenClaw), the power buttons, versions, backups and troubleshooting. Click Control Panel on your server card to get there.

Your AI model

Your agent answers with your own AI subscription or API key. There is no built-in free model, so your agent answers nobody until you connect one. You connect it inside the agent.

Connecting a provider inside your agent

On an OpenClaw server, the AI Model tab of the server page shows the card "Your AI model lives inside your agent" with two buttons: Open agent and Step-by-step guide. The steps inside the agent:

• Open your agent, go to Settings > Models and press Connect provider.
• For a ChatGPT plan (Plus or higher): first turn on Enable device code sign-in in ChatGPT under Settings > Security and login > App security. Then pick OpenAI > Codex login (device code).
• For Grok: pick xAI (Grok) > xAI OAuth. It needs SuperGrok or X Premium+.
• Or paste an API key for OpenAI, Anthropic, Google, OpenRouter and more.
• After adding a provider, pick it in the Default model dropdown.

On a Hermes Agent server, the AI Model tab holds the sign-in itself: "Use a subscription you already have" with ChatGPT, Grok, Nous Portal and MiniMax, or an AI Provider with an API key. See "Hermes Agent cards" further down this page.

For the full walkthrough, see Setting Up Your AI Provider.

Personality and instructions

Your agent's personality and standing instructions live in its workspace files. On OpenClaw you edit them in the Workspace Files card on the server page (or the Files link at the top of that page). SOUL.md holds the personality, IDENTITY.md the agent's name and role, USER.md what it should know about you, and AGENTS.md its helper agents. On Hermes Agent you edit the same kind of instructions inside your Hermes dashboard.

Good instructions usually include:

• A role definition. Tell the agent who it is. For example: "You are a friendly customer support assistant for an online pet store."
• Tone and style guidelines. Specify whether the agent should be formal, casual, humorous, or professional.
• Topic boundaries. Define what the agent should and should not discuss. For example: "Only answer questions about our products and services. Politely decline off-topic requests."
• Response format preferences. Tell the agent to use short answers, bullet points, step-by-step instructions, or whatever format suits your audience.

For a deeper guide with examples, see System Prompts & Personality.

Channels

Messaging channels such as Telegram, WhatsApp, Discord and Slack are connected inside your agent, which guides you through each one. The OneClickClaw dashboard has no channel form. On OpenClaw, open your agent and go to Settings, then Channels.

Channel status on the server page

On an OpenClaw server, the OpenClaw Agent card on the server page shows whether the agent is running and its gateway is connected, and lists your channels. Its buttons Manage channels in OpenClaw or Add a channel in OpenClaw take you into the agent. Mute disconnect alerts silences the alert when a channel drops. The Channels card further down shows the status of each channel.

Starting, stopping and rebooting

Your agent runs as long as its server runs. You control the server from the top of the server page or from the server card:

• Start: turns a stopped server on. Your agent starts with it.
• Stop: shuts the server down. Your agent stops answering until you click Start.
• Reboot: restarts the server and your agent with it. It usually takes a minute or two.

These buttons are hidden while an upgrade runs. On a Hermes Agent server, the Troubleshooting card can also restart just the dashboard or the messaging service.

Conversations and logs

Your conversations are kept inside your agent, on your own server. To read them, click Open Agent: OpenClaw shows them in its chat interface, Hermes Agent in its dashboard. The OneClickClaw dashboard does not show conversation content.

The Logs & Status tab of the server page shows the technical side: Bot Status, Gateway, AI Model, Version and Uptime, with the server logs and filters below.

Hermes Agent cards

A Hermes Agent server has cards of its own on the server page:

• Your Hermes Agent: the installed build and your dashboard sign-in. Email me a new password sends you a new password for the Hermes dashboard. Choose my own username and password unlocks when you keep your server on a paid plan.
• AI Model (its own tab): sign in with ChatGPT, Grok, Nous Portal or MiniMax, or choose an AI Provider and add an API key. After a sign-in, your server asks the model one short question before it says the AI is connected. Use Change Model and Apply Model to switch models later.
• Status card: shows whether your agent is answering, the secure connection and the installed version.
• Webhooks: turn on webhooks and create one, so other services can send events to your agent. Each request is signed with HMAC-SHA256 in the X-Hub-Signature-256 header. The limits are 30 requests a minute and 1 MB per request. The answer goes to the agent's log, Telegram or WhatsApp.

Tips for effective agent configuration

• Start simple, iterate often. Begin with short instructions and refine them based on how the agent responds in real conversations. You can change them at any time.
• Test on a private channel first. Before deploying to a public group, connect a private Telegram chat or a test Discord server. Verify the agent's tone, accuracy, and behavior in a controlled environment.
• Match the model to your needs. Larger models produce higher-quality responses but cost more per token. Smaller, faster models are better for high-volume, lower-complexity use cases. See Choosing the Right Model for a comparison table.
• Read your conversations regularly. Open your agent and check how it handles edge cases, off-topic questions and unclear requests. Use what you learn to improve its instructions.
• Use channel-specific instructions. If you connect Telegram, WhatsApp and Discord, note in your agent's instructions that it may be speaking to different audiences on each platform.