Browser analyzer

Run PromptScore directly in the browser

The browser analyzer on promptscore.dev uses the same deterministic scoring pipeline as the CLI, with built-in profiles that do not require filesystem access.

What the browser surface does

  • Parses the prompt in the client.
  • Runs the deterministic rule registry locally.
  • Builds the same `ScoreReport` shape used by the CLI and the Node library.
  • Uses built-in `_base`, `claude`, and `gpt` profiles instead of the YAML loader.
  • Leaves the hosted `promptscore.dev` analyzer offline by default, with no injected LLM client.

Browser-safe API

For client-side usage, import from `@promptscore/core/browser`. The browser entry exposes `analyzeWithProfile`, `format`, and the built-in profile helpers.

import { analyzeWithProfile, formatText } from '@promptscore/core/browser';

const report = await analyzeWithProfile('You are a helpful assistant. Summarize this article.', {
  profileName: 'gpt',
});

console.log(report.overall);
console.log(formatText(report));

Custom browser integrations can also pass `includeLlm: true` and an injected `llmClient`, but the hosted site does not enable that path today.

Privacy model

The current browser analyzer is local-first. Deterministic analysis does not send prompts to an external API.

Important: if you build a custom browser UI with an injected `llmClient`, keep that path clearly separated from the offline deterministic workflow so users always know when data would leave the browser.

When to use browser vs CLI

SurfaceBest for
Browser analyzerQuick checks, demos, onboarding, and embedding a local analyzer in a UI.
CLIAutomation, CI, scripts, and repeatable prompt linting in repositories.
Core libraryCustom integrations, internal tools, and app-specific workflows.
PreviousConfigNextRules Reference