Build something useful
Ship a working specialized AI agent or an agent-ready tool, integration, or toolkit. Your project should address a clearly defined problem or capability gap and demonstrate practical functionality.
Build specialized AI agents or the tools that power them.
Join a 48-hour hackathon focused on building practical AI systems. Create a specialized agent that solves a specific real-world problem, or develop tools, APIs, MCP servers, and integrations that extend what AI agents can do. Two tracks, one competition, and the freedom to build with the technologies you choose.
Agent Stack 48 is a 48-hour hackathon for developers building specialized AI agents and the tools that power them. From vertical agents solving industry-specific challenges to MCP servers, APIs, and toolkits that unlock new capabilities, the focus is on creating practical, working technology.
Ship a working specialized AI agent or an agent-ready tool, integration, or toolkit. Your project should address a clearly defined problem or capability gap and demonstrate practical functionality.
Use the models, frameworks, APIs, MCP servers, cloud services, and developer tools that make sense for your build.
One hour after kickoff, join the LabLab Discord for the live Q&A session and event support.
The next generation of AI isn’t just about smarter agents. It’s also about the tools that allow them to interact with the world.
Agent Stack 48 challenges developers to build either highly specialized AI agents designed for specific industries and workflows, or the tools and infrastructure that enable agents to access information, interact with systems, and perform meaningful actions.
Whether you’re building a vertical agent or a reusable toolkit, the goal is the same: solve a real problem, build something functional, and demonstrate its value.
Build an AI agent designed to solve a specific problem within a defined industry, profession, or workflow. Focus on depth, reliability, and practical results rather than general-purpose functionality.
Build tools that extend what AI agents can access, understand, or do. These can be standalone APIs, MCP servers, SDKs, connectors, integrations, or reusable toolkits that enable agents to interact with external systems.
Identify a specific workflow, industry challenge, or missing capability that your agent or tool will address.
Develop a specialized AI agent or build a tool, API, MCP server, or toolkit that enables new agent capabilities.
Show your solution in action through a working demo, real integration, or reproducible workflow.
Build with the models, frameworks, APIs, development tools, and infrastructure that best fit your project. Whether you’re creating a specialized agent or developing tools for other agents, there are no mandatory technologies or providers.
Explore Native.Builder to help build applications, interfaces, or components of your hackathon project. It may not fit every development workflow, and using it is entirely optional.
MCP servers, custom APIs, SDKs, and integrations are all valid project submissions, not just technologies used to build agents. Choose the architecture that makes the most sense for the capability you’re creating.
Start with a few useful resources, or bring your own stack. None of these tools are required.
Explore Native.Builder as an optional development tool for your project. New users receive 50 free credits to get started.
→ LabLab Tutorial Native.Builder Getting StartedA step-by-step LabLab guide to building and publishing with Native.Builder.
→ MCP Development Build tools for AI agentsLearn how to develop MCP servers, expose tools and resources, and connect AI applications to external systems.
→ Google AI Google AI StudioPrototype with Gemini models and create an API key for your project.
→ Google AI Gemini API DocsOfficial developer documentation for integrating Gemini into your application.
→ Google AI GemmaExplore Google's open model family and developer resources.
→ LabLab Tutorial LibraryBrowse LabLab tutorials covering models, APIs, agents, and developer tools.
→ Community LabLab DiscordJoin the community, get event support, and attend the live Q&A at 18:00 CET on Dec 11.
→All times are in Central European Time (CET).
The 48-hour build window begins immediately after the kickoff briefing. Start building from this point and continue until the submission deadline.
Bring setup, challenge, technology, and team questions one hour after kickoff. Join the LabLab Discord →
The build window remains open continuously. Develop, connect tools, test the workflow, and keep shipping.
The 48-hour build window ends. Complete and submit your project page, repository, demo, and submission materials by 17:00 CET.
Once your enrollment is approved, generate your personal referral link from the event dashboard, share it with builders, and watch your community score grow.
Referral points are a visible community score. They do not affect project judging or winner selection, and there is no separate referral prize pool.
Enroll in Agent Stack 48 and wait until your participation is approved.
Open your event dashboard and generate your personal referral link.
Invite builders and follow referral activity from the dashboard and live leaderboard.
One competition, two tracks, one set of winners. Whether you build a specialized AI agent or a tool that extends agent capabilities, your project competes for the same prizes. Full prize details will be announced before the hackathon begins. All project building must take place during the official 48-hour build window.
Details to be announcedA strong submission clearly demonstrates what you’ve built, the problem it solves, and how it works. Whether you’re submitting a specialized agent or a standalone tool, judges should be able to understand its purpose, technical implementation, and practical value.
An agent designed to solve a clearly defined problem within a specific industry, profession, or workflow. We encourage focused, purpose-built agents rather than general-purpose assistants or basic chatbots.
Yes. Tools and toolkits are an equally important part of Agent Stack 48. You can submit MCP servers, APIs, SDKs, integrations, connectors, or other solutions that extend the capabilities of AI agents. A complete agent application is not required.
No. Both tracks compete for the same overall prize pool. Projects are evaluated based on their technical quality, innovation, practical utility, and demonstrated functionality.
No. Agent Stack 48 is an open-stack hackathon. You’re free to use whichever models, frameworks, APIs, infrastructure, or development tools best suit your project. Native.Builder is available as an optional development resource, and new Native.Builder users receive 50 free credits.
No. MCP is one way to build tools for AI agents, but custom APIs, SDKs, direct integrations, and other architectures are equally welcome.
December 11 at 18:00 CET — exactly one hour after kickoff — in the LabLab Discord.
No. Referral points are a separate community leaderboard and do not affect project judging, prizes, or winner selection.
The prize structure will be added to this page before the hackathon begins.