Build Your Own AI Company: The Complete Paperclip Setup Guide

Visit Hostinger VPS now

Most people use AI as a simple chatbot. You ask a question, and it gives an answer. But what if you could build a whole company where AI agents hire each other, set their own goals, and work 24/7 without you touching a keyboard? That is exactly what Paperclip does. It moves you from managing a tool to presiding over a team of autonomous workers.

By the end of this guide, you will have a live AI organization. We will move from a blank server to a working company that researches tech news and sends a real digest to your inbox.

Understanding Paperclip: The AI Company Operating System

Paperclip isn't just another AI wrapper. It is a framework that lets you run a business via AI agents. In this setup, Paperclip is the company itself. It handles the org chart, the budget, and the rules. The AI agents are the employees.

Standard AI workflows require you to prompt every single step. Paperclip changes that. You set a high-level goal, and the agents take it from there. They wake up on a set schedule, check their task list, do the work, and then go back to sleep.

Launched in March 2026, Paperclip is an open-source project. This means it is free to use and you can host it on your own hardware. Because it is open-source, it grows fast and ships new features regularly.

Step 1: Deploying Paperclip on a 24/7 Server

You can run Paperclip on a laptop, but your company stops working when you close the lid. To have a business that works while you sleep, you need a Virtual Private Server (VPS). This ensures 24/7 uptime.

Hostinger is the top choice for this. They provide a one-click deployment template. This template includes a reverse proxy, which makes your connection safer over the web. You can use a discount link to save 10% on your setup.

Configuring the VPS and Deployment

When picking a plan, go with KVM 1. It has enough power to run Paperclip and a few other small services. Make sure to choose a 12-month term to keep your coupon active. You also get a free domain for the first year, which is handy if your agents eventually manage a website.

Once the server is live, you will see the configuration screen. Set your admin email and password. This is how you will log into your AI office. When asked for an AI provider, you can use an API key for models like Anthropic. However, for OpenAI, we will use a subscription method to keep costs low.

Step 2: Connecting the AI Engine (OpenAI Codex Authentication)

To make Paperclip work with your OpenAI subscription, you need to use Codex. This avoids paying for every single token via an API. But there is a critical first step.

Go to your OpenAI account settings. Find the security page and turn on "enable device code authorization for Codex." If you miss this, the login will fail and you will have to start over.

Using the Command-Line Login

Now, open the Docker Manager on your server and click "Terminal." This is where you talk to the computer with text. Follow these exact steps:

  1. Type docker ps and hit enter.
  2. Find the Paperclip container ID and copy it.
  3. Type docker exec -it [Your-ID] /bin/bash and hit enter.
  4. Notice the prompt change from root@server to node@. This means you are now inside the Paperclip project.
  5. Type codex login --device-auth and hit enter.

The terminal will give you a URL and a code. Open the URL in a new tab, sign into your OpenAI account, and paste the code. Once the terminal says "successfully logged in," you can close it.

Step 3: Launching and Structuring Your AI Organization

Now it is time to build the team. Open the Paperclip interface using the email and password you created earlier. You will go through four tabs to launch.

First, name your company. We will call ours "Tech Co." Give it a mission. For example: "Find the biggest latest tech news and provide the world's best daily digest via email." Keep this high-level. The AI uses this mission to make all its decisions.

Next, set up your CEO. Select the Codex adapter and run the "Test Now" button. A green "Passed" message means your AI engine is connected. Give the CEO a starting task, such as "Hire a founding engineer and create a hiring plan."

The Human as the Board of Directors

In Paperclip, you are not the manager. You are the Board of Directors. The CEO does the managing. When the CEO wants to hire a new agent, like a CTO, it sends a request to your Inbox.

You must approve these hires. This keeps you in control of your budget. If you want the company to be fully autonomous, you can toggle "Require board approval for new hires" to OFF in the settings. This lets the CEO build the team without asking.

Managing the Workflow

Your company is organized through an Org Chart and an Issues tab. The Org Chart shows who reports to whom. The Issues tab works like a Kanban board. You will see tasks move from "To Do" to "In Progress" and finally to "Done."

Each agent has a "Heartbeat" setting. This is how often they wake up to check for work. For a new company, set the CEO to 600 seconds (10 minutes). This speeds up the initial setup. You can increase this interval later to save money on tokens.

Step 4: Enabling External Tools for Real-World Output

AI agents can plan all they want, but they cannot reach the real world without tools. To send a newsletter, your team needs a way to search the web and a way to send mail.

Brave Search and Resend Integration

First, sign up for the Brave Search API. Choose the search plan. It gives you $5 in free credits every month. To avoid surprise bills, set a usage limit of exactly $5. This keeps the service free. Copy your API key.

Next, create an account at resend.com. This service handles the email delivery. Verify your email and generate an API key.

Managing Secrets in Paperclip

You don't want to paste raw API keys everywhere. Paperclip uses "Secrets." Go to the CTO agent's configuration tab. Under environment variables:

  1. Type brave_search_api as the key.
  2. Paste the Brave API key as the value.
  3. Click "Seal." This turns the key into a reusable secret.
  4. Repeat this for resend_api.

Now, if you want to give the Founding Engineer the same tools, you don't need to paste the keys again. Just toggle from "plain" to "secret" and select the existing secrets from the dropdown.

Step 5: Testing the Autonomous Workflow End-to-End

It is time to see if the company actually works. Instead of chatting with the AI, create a new "Issue."

Title the task "Test Tools" and assign it to the CTO. Give a clear instruction: "Use Brave Search to find tech news and Resend to email me a sample digest." Set the priority to "Critical."

You can watch the CTO work in real-time. The status will change to "In Progress." The CTO will search the web, write the summary, and trigger the email API.

Results and Delivery

After a few minutes, the issue will move to "Done." Check your email inbox. You should see a professional tech digest sent by your AI agents. Both the CTO and the Founding Engineer should be able to complete this loop.

For future work, assign big goals to the CEO. The CEO will then break those goals into smaller tasks and delegate them to the CTO or CMO. This saves you from micro-managing and lets the AI hierarchy handle the legwork.

Final Thoughts

You now have a fully operational AI company. It lives on a dedicated server, manages its own staff, and interacts with the real world through APIs. You shifted from being a prompter to being a board member.

The power of this setup is its scale. You can add more agents, give them more tools, and expand your mission. Whether it's a news digest or a complex research firm, your AI company is now live and working around the clock.

Post a Comment

Previous Post Next Post