Hey Product Hunt,

I built PullScope because PR review tools usually ask you to trust a backend before they help you understand risk.

PullScope takes a different route:

- paste a GitHub PR URL

- fetch public PR metadata and changed files in the browser

- score risk locally with deterministic rules

- optionally run AI review directly from your browser

The new angle I am most excited about is browser-native AI. PullScope now puts Chrome AI first: it checks Gemini Nano availability, can prepare the browser-managed model, and does not ask for a model API key in that path.

If Chrome AI is not available, PullScope still supports OpenAI-compatible BYOK endpoints such as OpenAI, Groq, Ollama, LM Studio, and custom providers.

What it is:

- a zero-backend GitHub Pages devtool

- local-first PR risk scoring

- browser-native AI review when available

- BYOK-compatible for external and local model endpoints

- raw model fallback so unexpected AI output is still visible

What it is not:

- not a SaaS

- not a proxy

- not a GitHub writeback bot

- not a place where model keys are stored on a server

I would love feedback from developers who review PRs daily:

- Is Chrome AI as the default path clear enough?

- Are the local risk signals useful before AI runs?

- What would make this worthy of being part of your review workflow?

Thanks for checking it out.