This track is strictly for framework-level work.

Build the best core extensions, improvements, forks, or entirely new open agent frameworks inspired by OpenClaw (or alternatives like ZeroClaw, NullClaw, etc.) and deployed on 0G. Focus on advancing how agents are created in 2026 - architectures, developer tooling, and infrastructure primitives that other builders will use.

Ideas to get you started:

🔹 New OpenClaw modules for hierarchical planning, reflection loops, or multi-modal reasoning that natively integrate 0G Compute’s sealed inference (live models like qwen3.6-plus or GLM-5-FP8) 🔹 Self-evolving agent framework that autonomously generates/tests/integrates new skills/tools using persistent 0G Storage memory 🔹 Modular “agent brain” library with easy swapping of memory layers (0G Storage KV/Log), LLM backends, or decision engines 🔹 No-code/low-code visual agent builder with one-click deployment to 0G Compute + Storage

This track is strictly for building the actual agents.

Create the most capable autonomous single agents, powerful multi-agent swarms/collectives, or any highly creative open agent project deployed on 0G. This track celebrates long-running goal-driven agents, emergent collaboration, and novel uses of iNFTs (ERC-7857) for ownership, composability, and monetization on 0G.

Ideas to get you started:

🔹 Personal “Digital Twin” agent that learns from user behavior and maintains evolving persistent memory via 0G Storage (KV for real-time state + Log for conversation/history) 🔹 Advanced research/knowledge or creative agent with persistent context on 0G Storage + self-fact-checking/reflection using verifiable 0G Compute inference 🔹 Specialist agent swarms (planner + researcher + critic + executor) that collaborate in real time via shared 0G Storage memory and coordinate inference on 0G Compute 🔹 iNFT-minted agents with embedded intelligence (encrypted on 0G Storage), persistent memory, dynamic upgrades, and automatic royalty splits on usage 🔹 Agent breeding/merging via iNFTs, emergent behavior experiments, agent arenas, or completely new paradigms leveraging 0G’s full stack

Build the future of agentic finance with Uniswap. Integrate the Uniswap API to give your agent the ability to swap and settle value onchain with transparency, composability, and real execution. Agents that trade, coordinate with other agents, or invent primitives we haven't imagined yet - we want to see it.

One prize pool, ranked. Best work wins regardless of what you build.

Build something on AXL, our peer-to-peer network node that gives your applications encrypted, decentralised communication with zero infrastructure. Your app talks to localhost, AXL handles encryption, routing, and peer discovery across the mesh. Any language that can make HTTP requests can use it.

We don't want to be prescriptive. If it works and it uses AXL in a meaningful way, it qualifies.

Please review the suggestions below on things you can build, and use your imagination:

• Agent Town: Build a multi-agent simulation where agents have distinct personalities, communicate freely over AXL, and interact in ways you can watch unfold. A generative-agents sandbox, a hackathon simulator, an agentic website with zero humans. Give agents personalities, let them loose, see what happens.
• Decentralised Agent Messaging: Build a messaging or discovery platform where AI agents find each other, form groups, and communicate peer-to-peer. Think Telegram, except every user is an AI agent. Agent social networks, agent marketplaces, self-organising working groups.

Or something else entirely!

All winners are fast-tracked into the Gensyn Foundation grant programme.

Judging Criteria: We want participants to create something on top of AXL with real utility, not just novelty. We will consider:

• Depth of AXL integration
• Quality of code
• Clear documentation
• Working examples

AI agents need persistent, human-readable identities too. Build a functional project where ENS is the identity mechanism for one or more AI agents. ENS should be doing real work — resolving the agent's address, storing its metadata, gating access, enabling discovery, or coordinating agent-to-agent interaction.

Most people know ENS for name → address lookups. We want to see what else it can do. Store verifiable credentials or zk proofs in text records. Build privacy features with auto-rotating addresses on each resolution. Use subnames as access tokens. Surprise us!

𝗧𝘄𝗼 𝗳𝗼𝗰𝘂𝘀 𝗮𝗿𝗲𝗮𝘀, 𝗼𝗻𝗲 𝗿𝗮𝗻𝗸𝗲𝗱 𝗽𝗿𝗶𝘇𝗲 𝗽𝗼𝗼𝗹. 𝗕𝗲𝘀𝘁 𝘄𝗼𝗿𝗸 𝘄𝗶𝗻𝘀 𝗿𝗲𝗴𝗮𝗿𝗱𝗹𝗲𝘀𝘀 𝗼𝗳 𝘄𝗵𝗶𝗰𝗵 𝗮𝗿𝗲𝗮 𝘆𝗼𝘂 𝗯𝘂𝗶𝗹𝗱 𝗶𝗻.

🟢 𝗙𝗼𝗰𝘂𝘀 𝗔𝗿𝗲𝗮 𝟭: 𝗕𝗲𝘀𝘁 𝗜𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝘃𝗲 𝗨𝘀𝗲 𝗼𝗳 𝗞𝗲𝗲𝗽𝗲𝗿𝗛𝘂𝗯.

Show us something we haven't seen before. Use KeeperHub's execution layer in a way that solves a real problem, whether that's an • agent, • a workflow, • a dApp, • a dev tool, • or something entirely new.

𝗪𝗵𝗮𝘁 𝗰𝗼𝘂𝗻𝘁𝘀: any project that uses KeeperHub (MCP server or CLI) in a meaningful way. No format restrictions. If it works and solves a real problem, it qualifies.

🟢 𝗙𝗼𝗰𝘂𝘀 𝗔𝗿𝗲𝗮 𝟮: 𝗕𝗲𝘀𝘁 𝗜𝗻𝘁𝗲𝗴𝗿𝗮𝘁𝗶𝗼𝗻 𝘄𝗶𝘁𝗵 𝗞𝗲𝗲𝗽𝗲𝗿𝗛𝘂𝗯.

Connect KeeperHub to something. Build the bridge so other developers don't have to. Two potential angles:

• 𝗣𝗮𝘆𝗺𝗲𝗻𝘁𝘀. Integrate KeeperHub with payment rails like x402 or MPP. Show how agents can pay for services, settle transactions, or route payment flows into KeeperHub execution.

• 𝗔𝗴𝗲𝗻𝘁 frameworks and tools. Build a plugin, connector, or SDK integration for ElizaOS, OpenClaw, LangChain, CrewAI, or any framework with an active builder community.

🟢 𝗝𝘂𝗱𝗴𝗶𝗻𝗴 𝗰𝗿𝗶𝘁𝗲𝗿𝗶𝗮: • Does it work? • Would someone actually use it? Real utility over novelty. • Depth of KeeperHub integration • Mergeable quality: clean code, clear documentation, working examples

🟢 A separate bounty for builders who give us honest, actionable feedback while integrating KeeperHub during the hackathon. Open to any team that uses KeeperHub, whether or not they place in the main prize pool.

We want feedback that helps us make KeeperHub better. • UX and UI friction, • reproducible bugs, • documentation gaps that slowed you down, • feature requests that would have made your build easier. This isn't a consolation prize. Honest feedback from builders working under time pressure is some of the most valuable product input we get.

𝗧𝘄𝗼 𝘄𝗶𝗻𝗻𝗲𝗿𝘀, $𝟮𝟱𝟬 𝗲𝗮𝗰𝗵.