Trust

Methodology

Most AI tooling lists are AI-generated listicles or vendors ranking themselves first. This index is built differently, and the process is open.

How a listing gets verified

A tool is added with its specs read from the official docs and repository, not from a marketing page.
Every entry carries a verified date and a verified by field naming the curator. 100% of the current set is marked verified.
License, open-source status, self-hostability, MCP support, and pricing are the facets that matter to practitioners and are checked first.

Status. The initial dataset is marked verified_by: seed, meaning best-effort data drawn from public docs. It is being progressively reviewed by hand. The freshness action keeps quantitative fields (stars, activity) current in the meantime.

What the novelty score means

Insights extracted from curated sources are embedded and compared against the rest of the corpus. A claim near 1.0 adds signal that almost no other source covers; a claim near 0.2 restates the consensus. High-novelty claims are surfaced first, and near-duplicate claims are deduped. This is the feature that lets us cut through the sea of repetitive reviews.

Freshness

A GitHub Action runs every Saturday and refreshes quantitative fields for every tool that has a public repository: star count, open issues, and last push. The knowledge pipeline runs on the same schedule to pull new transcripts and re-score novelty.

Editorial stance

No vendor self-ranking. A tool never writes its own entry.
Featured placement is disclosed and never influences the specs or insights shown.
The dataset is open: YAML and JSON in the repository.

The numbers right now

50 tools · 11 curated insights · 8 sources tracked.

Contribute

Spot something wrong? The data lives in src/data/. Open a pull request against tools.yaml, or file an issue, and a curator will review it.