Your first agent is ten minutes away

Getting started with Hivemeld does not require a technical background, a long setup process, or reading documentation for a weekend before anything works. The platform is built to get you from account creation to first agent output in a single session.

This post walks you through every step. By the end, you will have a live agent with a defined role, a system prompt, a Discord connection, and a completed first task. Not a demo. A real agent doing real work.

Step 1: Create your account and choose a workspace

Navigate to Hivemeld and create your account. The onboarding flow asks two questions: what kind of work do you do (professional, personal, or both), and what is the first problem you want to solve.

Your answer shapes the role recommendations the platform surfaces. If you say you want help with household management, it suggests a home management agent. If you say you want help with scheduling and email, it suggests an executive assistant agent. If you say you want help with content or research, it surfaces those roles.

You are not locked in by this choice. It is a starting point, not a constraint.

Step 2: Choose your first agent's role

Hivemeld is built around the concept that AI agents work best when they have a defined identity and scope. A general-purpose AI that does everything tends to do everything at a mediocre level. An agent with a role — and a clear understanding of what that role entails — does its specific job well.

The platform offers a library of pre-built role templates. For your first agent, pick the role closest to the problem you described in onboarding. Common starting points:

Executive Assistant — calendar management, email triage, scheduling, research
Meal Planning Agent — weekly or monthly meal plans, grocery lists, pantry tracking
Finance Agent — expense tracking, budget monitoring, recurring bill management
Home Manager — household inventory, maintenance scheduling, vendor management
Research Agent — topic research, summaries, competitive analysis, briefings

Select your role. The template loads a default system prompt that you will customize in the next step.

Step 3: Write your system prompt

The system prompt is the agent's operating instructions. It defines the agent's personality, its responsibilities, its communication style, and its decision-making boundaries. You do not need to write it from scratch — the template gives you a solid base — but personalizing it is what makes the agent actually useful to you.

The system prompt editor shows the template on the left and a preview of how the agent will introduce itself on the right. Customize it by adding:

Your specific context. If this is a meal planning agent, tell it about your household size, dietary restrictions, and cooking skill level. If it is a finance agent, describe whether you are tracking personal spending, business expenses, or both.

Your preferences for communication. How often do you want check-ins? Do you want daily summaries or weekly reports? Should the agent ask for confirmation before taking action, or act and report back?

Your decision-making boundaries. What can the agent handle autonomously? What should it flag for your review? For a first agent, err toward more checkpoints. You can loosen them as you build trust.

A good first system prompt is specific enough to be useful but not so detailed that it becomes brittle. Two to four paragraphs is the right range for most roles.

Step 4: Connect Discord

Hivemeld agents communicate through Discord. If you do not have a Discord server for personal use, creating one takes two minutes and is free. The platform walks you through creating a dedicated server if you need one.

Once your server exists, click "Connect Discord" in the Hivemeld setup flow. You will authorize the Hivemeld bot to join your server and be directed to create a channel for your agent. Name the channel after the agent's role — #meal-planner, #finance-agent, #home-manager — whatever makes sense to you.

The agent is now live in that channel. You will see it post a brief introduction confirming its role and what it is ready to do.

Step 5: Assign the first task

Do not overthink the first task. Pick something concrete and achievable — something where you can see the output and evaluate it easily.

For a meal planning agent: "Build me a meal plan for the next seven days. My household is two adults. We avoid shellfish and prefer high-protein meals. Weeknight dinners should take under 40 minutes."

For a finance agent: "Review my last 30 days of spending from the accounts I've connected and give me a breakdown by category. Flag anything that looks unusual."

For an executive assistant: "Draft a weekly schedule template for a Tuesday through Thursday in-office work week, with protected focus blocks in the morning and flexibility for calls in the afternoon."

Type your task into the agent's Discord channel and send it. The agent will acknowledge the task, confirm its understanding, and get to work.

Step 6: Read the first report

Depending on the task complexity, your agent will deliver its first output within a few minutes. For a meal plan, you will see a structured seven-day plan with notes. For a financial breakdown, you will see categorized spending with observations. For a schedule template, you will see a structured weekly view with rationale for the design choices.

Read it critically. The first output will not be perfect — it cannot be, because the agent is still learning your preferences. What matters in the first output is whether the agent understood the task and produced something structurally sound.

Respond with feedback directly in Discord. "The weeknight meals look good, but the weekend cooking times are longer than I have capacity for. Can you reduce all weekend meals to under 60 minutes?" The agent incorporates the feedback and revises.

This feedback loop is how the agent calibrates to you. Each correction narrows the gap between its defaults and your preferences.

What happens after day one

As described in Introducing Hivemeld, the platform is built around agents that improve over time. The first task output is a baseline. The fifth is noticeably better. The twentieth requires almost no correction.

The agent builds a model of your preferences from every interaction — what you approve, what you push back on, what you ask for repeatedly, and what you never ask for twice. It stores this context persistently, which means it carries what it learns from one task into the next.

After the first week, add a second task type. After the second week, consider adding a second agent for a different domain. Most Hivemeld users end up running two or three agents within the first month — not because the platform pushes them to, but because the first agent delivers real value and the obvious next problem becomes apparent.

The ten-minute promise

Account creation: two minutes. Role selection: one minute. System prompt customization: three minutes. Discord connection: two minutes. First task assignment: two minutes.

That is ten minutes from a blank browser tab to an AI agent actively working on your behalf. The first output is in your Discord channel before you have finished your coffee.

The rest is iteration. And iteration, with an AI agent, is the fastest path to a system that runs without you thinking about it.

Deploy your first agent now

Everything you need is already on the platform. Your first agent is ten minutes away.

Create your free Hivemeld account.

Deploy Your First Hivemeld Agent in 10 Minutes