This is a series of content I created for the Build with MeDo Hackathon at MeDo.
Introduction
In Indonesia, Jastip (concierge service) is very common. China has a similar culture called Daigou (buying on behalf).
The key similarity between Daigou and Jastip is one simple thing: "Access".
How Jastip works is basically someone from a different location with "access" to certain products offers them to others on social media. If people want to buy, the seller does not need to own the inventory or stock. They just need to be there at the store and ready to purchase. This Jastip service includes a service fee and sometimes delivery service. There are many kinds of products being sold, from snacks to luxury goods you cannot easily find in Indonesia.
People in Indonesia take this to another level by making it a full side quest while travelling. They plan a trip for fun, and to help finance the travelling, they do a Jastip side quest with many products, often from cross-country travel.
Then I found that China already has a similar culture with Daigou, but it is more focused on luxury items that are often hard to find in China or have some kind of rarity or reputation from the origin country (or so from what I read on the internet).
The interesting thing about China and Indonesia's Jastip/Daigou culture is how they operate. In China, Daigou mostly operates on WeChat livestreaming, and the storefront is built inside WeChat using mini programs.
On the other hand, in Indonesia, interaction and transaction usually happen on either WhatsApp groups or Instagram chat. It is very fragmented and causes multiple issues.
So to help my fellow Indonesians make money while travelling, I made an app to manage this kind of thing using medo.dev.
Show me the money
To motivate you to read further, let me put a motivational image for you.
Yes, the money. With this app, you can not only plan for travel, but also make some money with Jastip.
How, you may ask?
1. You can create a campaign
When creating a campaign, you set the starting date and end date of your Jastip.
But Jastip across countries is hard. You need to calculate currency, and I am also afraid of people doing hit and run.
Don't worry, this app handles the currency exchange automatically for you, and there is deposit payment for your buyer. They can use that deposit to buy from you later, but if they do not buy, you can redeem the deposit.
What's next after creating a campaign, you may ask?
2. You can open the campaign and check
Talk with the AI assistant so you are better informed about weather, currency, pricing strategy, adding products if you already have something in mind, or just reading the news in case something happens.
3. Talk with our smart AI agent
Our AI is not like ChatGPT or Gemini, where you need to give information about what you are doing, where, or when. It understands your Jastip context, and it is integrated deeply with MeDo's large language model.
You can also take notes for important information from the AI and check them from the notes tab.
4. Check who is already interested
Check your member tab to see who is already interested and has put down deposit money.
They actually pay with their debit/credit card (using MeDo Stripe skills). It will be yours to claim later.
5. Talk with your buyer directly
Instead of information scattered in WhatsApp and Instagram, you can interact directly here in the app. You can also send images if you want.
6. Lastly, download the invoice or just export everything into Excel for later.
You can enjoy your travel and manage the Jastip using our platform.
7. and heres from buyer side of the app
From the buyer side, the flow is browse campaign, join with deposit, send request, chat, and track the transaction.
Next is how did I make this app?
nd how did it only take me a week to do so...
1. Register to Medo.dev
Ok, MeDo is a full-stack AI coding platform. It sounds very much like a buzzword, but trust me, it is real. You may already have heard about v0, Lovable, Bolt, or Replit, but there is nothing like it. MeDo is more complete, and you only need MeDo (and Stripe apparently for payment).
2. Create new project
3. Next, you just wait until your idea is built
Here's the fun part. Medo.dev is actually not frontend first. It is product first. It will create the full product spec and requirements before it even starts coding. Everything is included, from what should be built, how the folders and system are structured, and what is out of scope, which is very important to make your product align with your vision.
You can edit it or not, it is up to you. Next, click generate the app and then wait.
4. PLUGINSSS (apparently now they call it skills)
The skill integrations in medo.dev blew my mind. In v0, Lovable, and Bolt, you are kind of forced to register for an outside backend service like Supabase. But medo.dev does this automagically, so you do not have to have a Supabase account and click connect or something like that. It just WORKS.
- login just works
- crud just works
- upload image just works
- realtime chat app just works
- even payment using stripe is just WORKS
And it is not only the backend. Even LLM, image generation, video generation, text-to-speech, and speech-to-text are included. No need to juggle multiple providers, grab API keys, store them, then integrate them. IT JUST WORKS.
You can just pay using a single credit system for all of the services. No need to set up multiple payments or subscribe to multiple websites to get API keys anymore.
I was really blown away by the depth of the integration, so I started to dig into the code.
5. The LLM skills
So in my experience using the medo.dev LLM skills, the default is not really great, but you can just prompt it. Make sure to always prompt it to make the output token to the max.
You can also ask it to call tools, using other skills as tool calls that the LLM can invoke. My biggest problem with the LLM skills is that it seems to be hardcoded to Gemini 2.5 Flash, while there are already newer models like 3 Flash or later.
You can actually make an agent in the LLM skills, not just a placeholder. For example, I built this AI assistant to be able to talk to its own CRUD API, so it can change the title at the top on the fly in realtime by calling directly to the internal API. Mind = blown.
Some tips and tricks using MeDo
Based on my few weeks using medo.dev, here are some tips and tricks that helped me boost productivity and save credits.
Basics
MeDo gives you an integrated backend and frontend workspace. You can look at the code, but you can also access files, logs, and infrastructure directly inside the app.
A single Fast Build prompt costs about 15 credits, while Deep Build costs 30 credits. As of writing this, new accounts usually get 300 credits on registration and 100 credits from daily login. That is around 10 Deep Build prompts or 20 Fast Build prompts.
Tip 1: Put related tasks in one prompt
From my experience, credits are counted per message, not by how hard the problem is or how many tool calls it uses.
There are limits if you ask for too many things at once, but if the tasks are related, it is usually better to put them in one prompt. This helps you get more value from each credit spend.
Tip 2: When debugging, ask for logs and checks
When something breaks, do not only ask MeDo to fix it. Ask it to add logs and checks first so it can understand the real problem.
Example prompt:
What is the best approach to fix this [issue]? Please add logging and checks to help identify the issue.
This helped me a lot when a problem was not fixed in one run.
Tip 3: Ask for design suggestions before giving strict direction
MeDo can be overly eager when you give it very strict design direction.
In my case, I gave it a reference and told it exactly what I wanted, but it still tried to be "creative" and the result was frustrating.
What worked better was asking for a few options first.
Example prompt:
Give me 3 design variants for this particular [problem]. The goal is to [specific goal], and I prefer it to look like [your preferences].
This way, MeDo gives you a few layout and style directions first. You can then choose one, adjust it, or ask for another variant.
This saved me a lot of credits because I stopped forcing one direction through many follow-up prompts.
Tip 4: when stuck, try to clear the context before sending a new prompt
Sometimes you get that little annoying bug, whether it is frontend or backend.
In my experience, it always helps to use the clear context button before sending another prompt. Clearing context can make the agent think and do the task better and faster.
this is the button, don't mind my negative credits i overused it lol
So that's what I am building and how I am building it, with tips and tricks.
