Methodology

How projects are discovered, selected and reviewed.

GitHub stars measure popularity, not quality or hands-on evaluation.

Updated 31 Oct 2026 · Star activity: September 2026.

Scope and discovery

Initial edition: 50 independently reviewed projects, ordered by GitHub stars; not a global ranking.

Most starred among indexed eligible candidates, not an exhaustive global ranking or quality rating.

  • Independent GitHub topic/text queries discover candidates; curated lists are not discovery sources.
  • Forks, archived and unrelated projects are excluded.
  • Ambiguous new projects require editorial eligibility/category/kind review.

Utility categories and kinds (agent, framework, platform, skill, infrastructure) answer different questions.

Top 50 and rising alternatives

Featured entries are ordered by current stars (up to 50).

  • Rising alternatives need at least 100 current stars and 50 stars created in the period, outside featured entries.
  • Within each kind, alternatives rank by absolute stars created, then current stars and stable ID.
  • Category allocation cycles through those kind groups, starting with their leading absolute counts; at most 3 alternatives, never forced.

Measuring star activity

Star activity: September 2026.

This edition selects nominal day buckets from 1 Sep 2026 up to, but not including, 1 Oct 2026.

  • Starred in September counts new star events, not net growth or percentages.
  • The history endpoint's Sunday-first day buckets are selected by nominal calendar date; underlying boundaries are not guaranteed UTC.
  • Missing/incomplete history is unknown, never estimated.

Source review and examples

Initial profiles were reviewed against primary READMEs and license documents, not hands-on runtime tests; this directory does not provide ratings.

  • Examples are illustrative and untested.
  • Profiles are evidence-based drafts, not claims of human authorship, endorsements, or tested runtime results.

Maintenance and retained profiles

Cached prose is retained across metadata refreshes; source changes are flagged for editorial review, not automatically rewritten.

  • GitHub archived=false does not establish active maintenance: check primary README notices.
  • The initial source review found AutoGen's README reporting archived/unmaintained status despite this metadata conflict; consult current upstream migration guidance before use.
  • Retained profiles remain publicly available outside current ranked lists; older metrics retain their original as-of dates.

Costs, privacy and licenses

A locally running process does not imply local-only data: external models, hosted services and telemetry can still apply.

Projects mix open-source and source-available licenses and product terms; check each primary license and provider policy rather than assuming unrestricted open-source use.

Data and schema

Directory JSON · Schema documentation