ToolVitals

What works

The core idea is genuinely useful: a health score for open tools backed by real receipts (commits, license class, release velocity, plus JSON/Markdown for agents). The agent-ready angle (llms.txt, public skill) is something almost nobody else does.

The biggest problem

Your score, which is the whole product, does not separate anything at the top. Supabase, VSCode, n8n, OpenCode all sit at 100/100 across every sub-metric. When every leader is a 100, the score stops meaning anything and I am back to just reading GitHub stars myself. Worse, your "Score Drops" list shows Framer, Calendly and Attio at 0. They are not dying, they are just not open source, so a quick reader thinks you are calling healthy companies dead. That kills trust in the rest of the data.

Top suggestion

Break the 100 ceiling so the score actually ranks (98 vs 94 vs 87 is useful, a wall of 100s is not), and pull proprietary SaaS out of the "declining" leaderboards entirely instead of scoring them 0. "Not open" and "in decline" are two different things and right now you are merging them.

What works

The agent-ready angle is the real differentiator and I'd lean on it harder. Markdown and JSON profiles per tool allows my agent to check "is this dependency dying" mid-plan without me tabbing out. 3,392 tools tracked with deterministic signals is a solid dataset. That's the part I'd build on.

The biggest problem

I can't audit the number I'm supposed to bet my stack on. Four dimensions (Health, Shipping, Source Coverage, Confidence) roll into one score and the homepage never says how, and the evidence trail lives on a separate methodology page instead of on the tool profile where the decision happens. A score I can't trace to commits and releases is just a vibe. Adopting a dependency is a decision I'd like to see receipts on before making.

Top suggestion

Put the evidence on the profile: score, then the three signals that moved it most, each linking to the raw event (last release, commit trend, license change). And ship the agent story properly: a documented JSON endpoint with provenance fields. I'd wire that into my planning tool's dependency checks tomorrow, and that's the version I'd pay for.