Getting the Most From Your AI Agent

First: A Quick Win to See the Power

Before diving into setup, let's set up one thing that'll save you 15 minutes every morning — a Daily Briefing.

That's one example. Your agent can handle research, reminders, email drafts, file organisation, data lookups — anything you'd normally do manually at a computer. The key is teaching it about you first.

0. Before You Start: Getting Help

If you get stuck at any point, you have two options:

• Ask your agent. Seriously — type "How do I set up daily memory?" and it will help you.
• Check the health of your agent by visiting your domain. If the chat loads, you're good. If it shows "starting up", wait 30 seconds and refresh.

1. Teach It Who You Are

Right now your agent is generic. It doesn't know your name, your work, or how you like things done. Fix that by updating three files in your agent's workspace:

• USER.md — Who are you? What do you do? What's your timezone? What matters to you?
• IDENTITY.md — Who is the agent? Give it a name, a role, a purpose. "You are Atlas, a business operations assistant for a property management company."
• SOUL.md — How should it behave? Formal or casual? Detailed or concise? Should it be proactive or wait to be asked?

Why this matters: Without this, your agent gives generic answers. With it, every response is tailored to you — it knows your context, your preferences, and your priorities.

2. Make Sure It Remembers Things

Your agent comes with QMD — a structured memory system — already installed and active. Think of it as your agent's long-term brain. It doesn't just remember the current conversation — it remembers things across days, weeks, and months.

QMD uses a file called MEMORY.md to store important things it learns about you and your work, plus daily notes that capture what happens each day.

What to do: Tell your agent:

• "Make sure you maintain your memory file. Every day, create a daily note. When you learn something important about me or my work, save it to MEMORY.md."

What good memory looks like:

• It remembers your clients' names after you mention them once
• It remembers decisions you've made ("Last week you decided to go with Vendor B")
• It learns your preferences over time ("You prefer bullet points over long paragraphs")

3. Your AI Model (Already Done)

Your agent comes with a pre-configured AI model and $5/month of free usage — enough for hundreds of conversations. You don't need to do anything here unless you want to upgrade.

Optional upgrades (visit /settings to change):

• Claude by Anthropic — Significantly smarter for complex reasoning and writing. Requires your own API key (~$5–20/month depending on usage).
• Your own Groq key — Same model, no monthly limit. Free to get at console.groq.com.

4. Keep Your Stuff Safe

Your agent runs on its own private server — nobody else has access. But there are a few good habits:

• Don't paste sensitive passwords directly in chat. Use the /settings page for API keys.
• If you connect Telegram: Set up an allowlist so only you (and people you choose) can talk to the agent. Ask your agent: "Only allow messages from my Telegram account."
• Your gateway token is already set up. This is the key that keeps your agent's web interface private.

5. Connect Telegram (Optional but Powerful)

The web chat works great, but Telegram turns your agent into something you can message from your phone — like texting a very smart assistant.

To set up:

1. Open Telegram, search for @BotFather
2. Send /newbot, follow the prompts, get your bot token
3. Tell your agent the token in chat — say "My Telegram bot token is [paste token]" and it will configure itself
4. Message your new bot on Telegram — your agent will reply

6. Give It Web Access

By default, your agent can't browse the internet. Adding a Brave Search key (free) lets it:

• Look up current information (news, prices, weather, company info)
• Research topics and summarise findings
• Fact-check and verify things before giving you answers

To set up: Get a free key at api.search.brave.com, then add it in /settings.

7. Set Up Recurring Tasks

This is where your agent becomes a real time-saver. It can run tasks on a schedule — every morning, every Friday, every hour — without you asking.

Ideas for recurring tasks:

• Morning briefing — News + email summary + today's priorities
• Weekly report — Summarise the week's activity every Friday at 5pm
• Price monitoring — Check competitor prices daily and alert you to changes
• Social listening — Monitor mentions of your brand or industry keywords
• Email triage — Flag urgent emails and draft responses for routine ones

To set up: Just tell your agent what you want and when. For example:

8. Build Your Own Shortcuts

When you find yourself asking your agent to do the same thing repeatedly, turn it into a skill — a saved workflow it can run with a single command.

Examples:

• "Summarise this" — Takes any text, article, or document and gives you the key points
• "Draft a reply" — Takes an email and writes a professional response in your style
• "Research [topic]" — Searches the web, reads multiple sources, gives you a structured brief

Once everything is checked, your agent isn't just installed — it's production-ready and saving you time every day.