Discer.io

Learn Agentic AI through play. Discer.io is a MMO sandbox where you build and deploy agentic AI workflows using drag-and-drop blocks—like Scratch, but for AI agents.

Comment 3

"Discere" - to learn

Agentic AI literacy is broken. Most people have heard of agents but few actually understand how to build them. There’s no accessible, game-like way to learn how multi-step reasoning, tool orchestration, and feedback loops come together to form a working agentic system. Current resources are either text-heavy or code-intensive, leaving beginners overwhelmed and disengaged.

Our Solution

We built Discer.io, an educational MMO sandbox that lets players design, orchestrate, and deploy agentic AI workflows using drag-and-drop programming blocks and prompt-based commands. Think Scratch, but for agentic AI systems.

It turns abstract AI workflows into tangible, interactive experiences. So Instead of just reading about agents or watching demos, you’ll build your own, connect tools, set reasoning steps, and see the workflow come alive — all without writing a single line of code.

How We Built It

We combined Next.js for the frontend, Python for the backend, and a custom tick-based simulation environment built using bun, to bring agent logic to life. The visual programming was powered by modular blocks linked to custom built AI reasoning modules. Also, we utilized SnowflakeSQL for robust log and data extraction, and deployed the services using Digital Ocean. We used Dedalus Labs for our agentic capabilities and MCP server hosting. Our agents may use any of a variety of models, including Grok, Gemini, Claude, and GPT.

Challenges We Faced

Balancing educational clarity with open-ended creativity was tough. We had to design a workflow system that feels like play yet mirrors real agentic reasoning. Furthermore, building an entire video game as one small component of a larger product, pushed us to think outside the box and solve problems with maximal efficiency.

What We Learned

We learned how to visualize multi-step reasoning in real time, turning invisible thought chains into real, interactive systems. Thinking from the lens of an educational product also presented a unique opportunity to design an engaging curriculum and develop a service for people of all ages and backgrounds.

What’s Next

We plan on expanding multiplayer features, classroom integrations, and open-source “agent packs” so anyone can remix and learn from others’ designs. Additionally, we would love to see a wider array of scenarios, that push the learners and agents in creative and fun ways.

Built With

Share this project:

Updates