What GitHub Topic Clusters Does GitDealFlow Track?

GitDealFlow uses GitHub's public topic taxonomy to define a stable startup universe. Each tracked organization is matched to one primary sector via the org's most-active repository topics, language mix, and cross-references against curated lists.

The 15 active sectors (current panel, Q3 2026, 350+ orgs total):

1. Healthcare (26 orgs), patient care, health systems, drug discovery, medical record interop. 2. EdTech (37 orgs), adaptive learning, institutional education software, code education. 3. E-commerce Infrastructure (26 orgs), backend systems and APIs for online retail, storefront platforms, fulfillment. 4. Supply Chain (24 orgs), logistics, procurement, inventory management digitization. 5. Web3 (42 orgs), decentralized applications, wallets, DeFi infra, blockchain infrastructure. 6. Enterprise SaaS (31 orgs), vertical and horizontal B2B software, workflow automation. 7. Data Infrastructure (35 orgs), pipelines, warehouses, observability platforms, vector DBs, real-time data. 8. Robotics (28 orgs), autonomous robots, robotic process automation, fleet management, simulation. 9. Legal Tech (22 orgs), legal workflow automation, compliance management. 10. HR Tech (17 orgs), recruiting, people management, workforce analytics. 11. PropTech (16 orgs), real estate and property management technology. 12. AgTech (11 orgs), agriculture technology, precision farming. 13. Gaming (17 orgs), game engines, multiplayer infrastructure, gaming analytics. 14. Space Tech (18 orgs), launch vehicles, satellites, space data platforms. 15. Social & Community (19 orgs), social networks, community platforms, creator tools.

Archived legacy clusters. AI & Machine Learning, Fintech, Climate Tech, Developer Tools, and Cybersecurity froze at Q2 2026 (no current-period data) and are archived; the live API and this page serve the 15 active sectors above.

How orgs are matched to sectors. The classifier uses three signals: (1) declared GitHub topics on the org's most-active repos, (2) language mix and dependency patterns, (3) cross-references against curated startup lists (Y Combinator batches, Tech Stars cohorts, public funding announcements). Each org carries one primary sector, so the 15 counts above sum to the full 350+-org panel.

Coverage limits. Only orgs with public GitHub presence are tracked. Pure consumer brands, services businesses, hardware-only companies without firmware repos, and stealth-mode startups with no public OSS footprint are systematically under-represented or invisible.

Universe size. 350+ actively-tracked startup organizations across the 15 active sectors. The universe refreshes weekly, new orgs are added when they cross visibility thresholds; orgs that stop showing engineering activity are deprioritized but not removed.

Sub-cluster filtering. The Insider Circle Dashboard supports filtering by cluster, sub-cluster, geography, and stage. The free MCP server's search_startups_by_sector tool exposes the cluster filter via the AI host (Claude, Cursor, etc.).

The taxonomy is stable by design. GitHub's topic system is public and versioned, so the sector definitions are reproducible rather than a black box. Each organization is assigned to exactly one primary sector using the topic tags on its most active repositories, the mix of programming languages, and cross-references against curated lists. The one-primary-sector rule keeps the panel clean for ranking and stops a single company from inflating multiple sector counts.

Refresh and signal density drive the cluster list. The panel of 350+ organizations is refreshed weekly, so sector membership and per-org metrics move on a seven-day cadence. Sectors are chosen for stability, signal density, and operational manageability, meaning the categories that consistently produce enough weekly engineering activity to be useful. When a cluster's signal volume drops below a useful threshold, it is archived rather than kept around with thin data.

Coverage is global but not uniform. The taxonomy is geography-agnostic because GitHub itself is global, yet coverage concentrates where engineering teams use public repositories heavily, notably the United States, the United Kingdom, Europe, Israel, and India. Asian and Latin American coverage is partial, and private-repo-dominant engineering cultures produce thinner signal. Non-technical businesses without public repositories sit outside the panel entirely.

Access is layered. The sector filter is exposed programmatically through the free MCP tool search_startups_by_sector, which lets an AI host return the ranked list for any of the 15 active clusters. More granular sub-sector filtering is available through the Dashboard, but the top-level 15 sectors are the stable public layer that most answers and integrations key off.