Elasticsearch Agent Builder Hackathon

About the challenge Welcome to the Elasticsearch Agent Builder Hackathon, running from January 22 to February 27, 2026. Build context-driven, multi-step AI agents that combine search, workflows, tools, and reasoning to automate real-world tasks. This isn’t about single prompt answers, it’s about agents that retrieve the most relevant context, select the right tools, and execute tasks across multiple steps. Your mission is to design agents that actually get work done. Compete for a share of $20,000 in prizes, collaborate with Elastic engineers, and get featured on our blog and social channels. We can’t wait to see the agents you develop Why participate ? Build agents for real work: Design agents that reason, use tools, keep context, and actually finish tasks. Level up fast: Practice multi-step reasoning, tool orchestration, and stateful workflows under real constraints. Win and get seen There is $20,000 on the table, plus exposure on blogs and social channels. What's Elasticsearch and Agent Builder Elasticsearch is an open source, distributed search and analytics engine built for speed, scale, and AI applications. As a retrieval platform, it stores structured, unstructured, and vector data in real time. It delivers fast hybrid and vector search, real-time analytics on logs, metrics or events, and enables AI-driven applications with high performance, accuracy, and relevance. Agent Builder is a framework to rapidly create custom agents that connect LLMs to private Elasticsearch data. Tailor tools to query indexes, trigger workflows, or use piped query language for logic, relevance, and personalization. It supports custom agents accessible via Kibana, APIs, MCP, and A2A. Requirements What to Build Build a multi-step AI agent that uses Agent Builder combining a reasoning model with one or more of its available tools (Elastic Workflows, Search, or ES|QL) to automate real-world tasks. There are no category restrictions. Below are some strategic tracks for inspiration and check out the resources tab Automate messy internal work: Agents that handle real, brittle workflows like support triage, sales ops, compliance checks, data cleanup, or incident response . Build tool-driven agents: Agents that actively use APIs, SDKs, and integrations, and choose the right tool at the right time instead of relying on prompts alone. Narrow agents built for one domain (healthcare, fintech, legal, logistics, marketing, dev tools) that solve a clear, real problem. Show measurable impact: Agents that clearly demonstrate time saved, steps removed, or errors reduced, even with rough numbers. Multi agents: Applications where different agents plan, execute, review, and verify work, including resolving disagreements. Embed agents where work already happens: Agents that live inside Slack, email, IDEs, ticketing tools, or internal dashboards and feel native to existing workflows. Connect disconnected systems: Agents that glue together tools that don’t normally integrate, like CRM + support, GitHub + CI, or calendar + email. Time-series and geo aware agents: Agents that detect patterns across time and location. Spotting anomalies, trends, or events in metrics, logs, or sensor data, and taking action based on when and where something changed. Let agents take reliable action: Agents that safely execute decisions creating tickets, deploying fixes, updating records—and explain what they did and why. What to Submit Brief description describing the Agent that was developed (~400 words). Must include the problem solved, features used, and 2-3 features you liked or challenges you had. ~3-minute demonstration video Provide a URL to your open source code repository for judging and testing. The code repository must be public with an approved OSI Open Source License Get extra points - share a post your project about Agent Builder on any social channel and include the post link in your submission or tag @elastic_devs or @elastic on X