Our first hackathon: Partnering with ETHGlobal for OpenAgents
We've partnered with EthGlobal to run our first week long online hackathon on OpenAgents. It's our first hackathon, and we couldn't have picked a better one if we'd tried.
Seven years of unglamorous work
KeeperHub didn't start as a product. It started as the infrastructure we kept rebuilding for protocol clients who needed transactions to land. The job was never glamorous. Smart gas estimation. Exponential backoff. Private routing. Mempool simulation. Audit logs that would actually hold up if someone needed to reconstruct what happened at 4am on a Sunday.
For a long time, this work was invisible. Nobody writes a Twitter thread about retry logic. There's no conference track for "the boring middle layer between your trigger and your transaction." But the protocols whose TVL we were responsible for cared about exactly two things: did it execute, and can we prove what happened. Everything else was theater.
So we kept building. Through congestion events. Through gas spikes that took out half the keepers in the ecosystem. Through Black Thursday. Through every protocol upgrade and every chain split.
Most of those years, we were a service. Protocols hired us, we kept the lights on, the work disappeared into the ops layer of someone else's product.
The shift from service to platform
About two years ago, we started turning the engine into something other teams could use directly. The same retry logic, the same gas optimization, the same audit trail, but exposed through interfaces that didn't require a six-month onboarding from our engineers. A workflow builder. A CLI. Then a Model Context Protocol server. Then x402 integrations (with MPP coming soon). Then plugin support so teams could extend it themselves.
The bet was simple: if we did this work for the biggest protocols in DeFi, smaller teams would want it too. They just needed a way in that didn't involve a Statement of Work and a Slack channel.
The agent execution gap, in plain terms
Sometime in the last 18 months, AI agents stopped being a research curiosity and started being a real category. ElizaOS, AgentKit, LangChain, framework after framework, each of them solving the reasoning side of the problem. Better planning. Better tool use. Better task decomposition. Agents that can reason about markets, score yields, decompose multi-step strategies.
And then they all hit the same wall. The agent decides to act. Now what?
The transaction needs to land. Gas conditions change between the decision and the broadcast. The mempool might be hostile. The retry logic, if it exists at all, is somebody's weekend project. There is no audit trail. There is no SLA. There is no human on the other end of the wire when something silently dies at 3am.
This is the agent execution gap. It's the unglamorous middle that turns a clever agent into a production agent. And it's exactly the problem we've been solving for protocols since before "AI agent" was a category anyone took seriously.
Why OpenAgents, and why now
ETHGlobal runs hackathons on themes that reflect where web3 is actually moving. Last year it was scaling and proofs. This year, one of the flagship events is OpenAgents, a hackathon entirely about autonomous agents on Ethereum. Agent frameworks. Onchain identity. Agent payments. Agent execution.
For us, this is an unusually clean alignment. We didn't have to pitch the organizers on the relevance of execution infrastructure to AI agents. The relevance is the entire premise of the event. We've been building the answer to a question that, this April, is going to be asked by a few hundred of the best builders in the space, all in one room, on a deadline.
What we're bringing to the room
We're partnering with ETHGlobal for Open Agents with $5,000 in prizes across three tracks. Each track has its own winner. We'll be sharing the full prize details, judging criteria, and workshop info as we get closer to the event. Follow along on X and LinkedIn.
The first of many
This is our first hackathon as a named partner. It will not be our last.
Hackathons are where that conversation actually happens. Not in pitch decks. Not in landing page copy. In a room full of builders, on a clock, trying to ship something that works.
If you're interested in how the agents on chain space has been developing then make sure that you apply before April 23. The hackathon runs April 24 to May 3, finalists will be announced on May 6, and we'll be there the whole way.
See you at OpenAgents.
