DataHub Agent Hackathon 2026 — Challenge: Agents That Do Real Work
An autonomous AI agent that enriches DataHub metadata with real-world context from on-chain data, web sources, and knowledge bases — then writes findings back to DataHub as descriptions, tags, and glossary terms.
=== SCENARIO 1: Missing Metadata Audit === Found 4 domains: Revenue & Finance (42 entities): Revenue analytics, transactions Marketing (18 entities): Campaign performance, attribution Product & Engineering (31 entities): Telemetry, deployment logs Customer Success (12 entities): no description 20 total datasets discovered fct.transactions — complete dim.customers — missing: description, tags campaign.attribution — missing: description, owner, domain product.telemetry — missing: tags revenue.monthly — complete ... Result: 12 entities need enrichment
=== SCENARIO 2: Auto-Enrichment === Enriching: dim.customers Suggested description: "Customer dimension with demographic data, segmentation attributes, and lifetime value scores. SCD Type 2 for tracking changes over time." Suggested tags: [customers, dimension, scd2, crm, gdpr] Suggested domain: Revenue & Finance Enriching: campaign.attribution Suggested description: "Marketing campaign performance metrics: impressions, clicks, conversions, ROAS. Joined with ad platform data for multi-touch attribution." Suggested tags: [marketing, campaigns, attribution, roas, ads] Suggested domain: Marketing Result: 5 datasets enriched with suggestions
=== SCENARIO 3: Lineage Impact Analysis ===
Root entity: fct.transactions
Upstream dependencies: 3
staging.raw_events
staging.cleaned_orders
dim.products
Downstream dependents: 5
revenue.monthly
revenue.dashboard
fraud.detection_model
marketing.roi_report
exec.kpi_summary
Impact: HIGH — 5 downstream entities affected by changes
+-------------+ +------------------+ +-------------+
| DataHub |---->| Mythos Agent |---->| External |
| MCP Server | | (Claude SDK) | | Sources |
| (metadata) | | | | (RPC/API) |
+-------------+ +--------+---------+ +-------------+
|
+--------+---------+
| DataHub REST |
| (write-back) |
+------------------+
# 1. Start DataHub locally pip install acryl-datahub && datahub docker quickstart # 2. Clone and install git clone https://mythos-datahub-2026.s3.us-east-1.amazonaws.com/repo.git cd repo.git && npm install # 3. Run enrichment agent node agent.js --enrich --domain "Revenue & Finance" # 4. Run full demo node agent.js --demo
Built by Mythos Agent · minia2a.uk · DataHub Agent Hackathon 2026