How the index works
A ranking is only useful if you know how it was made. This page is the complete methodology — what gets listed, how it's classified, how it's ranked, and where the money comes from.
What gets listed
The index covers working software in the AI-agent ecosystem: agents, agent frameworks, platforms, MCP servers and clients, SDKs, and the infrastructure they run on. Every listing is assigned exactly one primary type, shown on its row and page.
We exclude tutorials, courses, curated link lists, and repositories that merely mention agents or MCP without being part of the ecosystem. Our audits have removed 62 entries on these grounds so far — including a music player and an nginx admin panel that had tagged themselves into agent categories on GitHub.
How open-source rankings work
Open-source listings are ranked by maintained adoption: GitHub stars as the primary signal, qualified by maintenance activity (last release activity is shown on every row). Stars are an imperfect proxy — which is why we audit them.
Repositories showing star-farming patterns — implausibly sustained star velocity on young repositories from unknown publishers — are delisted entirely, not down-ranked. Our audit heuristic flags anything above roughly 100 sustained stars per day on repositories under 18 months old; every flag is reviewed by hand before action, so genuinely viral projects from recognized publishers stay.
Why commercial products are not ranked by stars
Commercial platforms don't compete on GitHub stars, so comparing them to open-source repositories by star count would be false equivalence. Commercial products appear in a separate, alphabetical section on category pages. We show pricing model and product facts instead of repository metrics.
Sponsorship policy
Sponsored placements are how the index stays free. The rules are strict: a sponsored listing never receives a rank number, never appears inside the numbered organic index, and is always labeled Sponsored. The first numbered result on any category page is always the top organic result.
Some outbound links are affiliate links, which may earn us a commission at no cost to you. Affiliate status never affects classification, ranking, or editorial content.
Descriptions and editorial content
Open-source listing descriptions are extracted from each project's own README and labeled as such — we don't generate marketing copy on a project's behalf. “Best for / Avoid if” guidance is written by us, only for tools we can describe with confidence, and reflects editorial judgment, not sponsorship.
Freshness and verification
Repository data (stars, activity, language, license) is refreshed from the GitHub API, and every listing shows its “data verified” date. Stale or archived projects are removed as they are found.
Corrections
Wrong classification, outdated facts, or a project we should not have delisted? Email corrections@agentsearchengine.app or use the submission form — corrections are reviewed by hand and applied to the live index. Press: press@agentsearchengine.app.