Enterprise DNA
Build Index / Methodology

Methodology

How our agents score what to include.

The scoring rubric our agents use to decide what makes Build Index.

Every candidate the crawler discovers gets routed through an editorial agent that scores it against the rubric below. Only entries scoring 70+ across the rubric ship. The rubric is shared across the six directories, but each directory weights it differently.

The base rubric

  1. Real, not announced. Is this a product that exists and that a non-trivial team is using, or is it a tweet plus a waitlist?
  2. Deployable today. If a reader wanted to use this within an hour, could they? We rate effort from one-click to high and we are unforgiving.
  3. Trajectory. Is the vendor shipping, is the GitHub repo active, are the docs getting updated? We deprioritize abandoned projects even if they were once excellent.
  4. Differentiation. Does this do something the other options don't? Cool features, novel UX, an unusually clean integration story.
  5. Coverage. Does this fill a category gap? We will include a slightly weaker entry if it represents a category we otherwise have no representation in.

Directory-specific weighting

  • Agents weight Trajectory and Differentiation. A coding agent that shipped six features this quarter beats one that shipped a benchmark.
  • Skills weight Real (does the skill actually fire) and Deployability (drop in, done).
  • MCP Servers weight Deployability and Coverage (the protocol is young, gaps matter).
  • Open Source weights Trajectory hard via GitHub recency and stars, with a real-world-use bonus.
  • Apps and SaaS weight Real adoption signals and Differentiation over polish.
  • Builder Showcase weights Real builds, real builders, and the breadth of cross-references back to the rest of the index.

Cross-references are the moat

Every entry declares relations forward: what platforms it works in, which MCP servers it uses, what it's built with, what it pairs well with, what it's an alternative to. The build computes the reverse. That makes the index queryable structure: an MCP server's page lists every agent that uses it. A framework's page lists every showcase that ships against it. Click any link and you keep traversing the graph.

How often we refresh

Every entry is re-scored monthly. Vendors that stagnate get demoted from featured. Vendors that ship aggressively can climb back. We log every refresh in the changelog.

What we don't do

  • We don't take payment for placement. Every featured slot is editorial.
  • We don't include things we haven't read or run. If we couldn't get the tool to do anything useful, it doesn't ship.
  • We don't fluff up entry counts. Quality first, breadth second.