CertScore.ai

CertScore

CertScore (certscore.ai) is a production-minded MVP for scanning public websites for potential accessibility, privacy, cookie, policy, and disclosure risk signals. It is a risk signal and monitoring product, not a legal certification platform.

Monorepo structure

website-signal-risk-scanner/

├─ apps/ │ ├─ web/ │ └─ validation-worker/ ├─ packages/ │ ├─ shared/ │ ├─ web-bot-auth/ │ ├─ db/ │ └─ ui/ ├─ docs/ ├─ .env.example ├─ turbo.json ├─ package.json └─ pnpm-workspace.yaml

Workspace packages

• apps/web: product-facing web app and control-plane workflows
• apps/validation-worker: active validation runtime owned by WC01
• packages/shared: shared constants, types, validators, scoring config, and scheduling helpers
• packages/web-bot-auth: server-only Web Bot Auth signing and key-directory helpers
• packages/db: PostgreSQL query helpers, migrations, seed SQL, and env helpers
• packages/ui: reusable UI primitives

Repo boundary

WC01 is now the product/control-plane repo.

• product web flows, scan creation, reporting, and validation stay here
• the standalone scanner runtime now lives in WS01
• scanner operational changes, crawler identity work, and scanner deploy flow should originate in WS01

What the MVP includes

• public homepage with preview scan funnel
• Better Auth with Google OAuth and email/password login flows
• organization bootstrap and protected workspace routes
• domain management, plan limits, and client grouping
• DB-backed scan creation with scanner-service claiming
• crawl/discovery, accessibility, privacy, and legal heuristics
• deterministic scoring, canonical report payloads, and authenticated report UI
• scanner-generated PDF generation and upload
• regression summaries and scheduled rescans
• lightweight branding for reports and PDFs

MCP Light

CertScore.ai MCP Light is the free, no-auth, hosted three-tool integration for public website privacy scans. It detects evidence-backed privacy, cookie, tracker, consent, policy, regulatory-review, HTTPS/TLS, and eligible Reject Path post-refusal signals. Results include a CertScore score and supporting evidence for human and agentic review; they are not legal advice, certification, or a compliance determination.

Add CertScore to Cursor · MCP Light landing page · Cursor Directory listing · Agent installation guide · Full installation reference · MCP package guide

Try it with: Use CertScore.ai to review [PUBLIC URL] before launch. Report the CertScore score, evidence-backed findings, Reject Path observations when available, coverage limitations, and the report URL.

Environment variables

This monorepo should use apps/web/.env.local as the single local development runtime env:

• copy apps/web/.env.example to apps/web/.env.local

Use .env.example only as a reference template for shared keys. Do not rely on a root .env.local for app runtime configuration.

Recommended environment split inside WC01:

• local web + local validation runtime: dedicated dev PostgreSQL database plus S3-compatible storage
• production web + production validation runtime: dedicated production PostgreSQL database plus S3-compatible storage

Do not point localhost at the production database or production auth credentials unless you are intentionally testing production behavior.

Required for the web app:

• NEXT_PUBLIC_APP_URL
• DATABASE_URL
• BETTER_AUTH_SECRET
• GOOGLE_CLIENT_ID
• GOOGLE_CLIENT_SECRET
• S3_BUCKET
• S3_REGION
• S3_ACCESS_KEY_ID
• S3_SECRET_ACCESS_KEY

Required for the validation runtime in WC01:

• DATABASE_URL
• OPENAI_API_KEY

Optional but recommended:

• WORKER_CONCURRENCY
• PLAYWRIGHT_BROWSERS_PATH
• VALIDATION_OPENAI_MODEL
• VALIDATION_NANO_MODEL
• WEB_BOT_AUTH_ENABLED
• WEB_BOT_AUTH_PRIVATE_KEY_PEM
• WEB_BOT_AUTH_SIGNATURE_AGENT_URL

Local development setup

Use Node 22, 23, or 24 for local development. Node 20 is below the workspace engine floor, and Node 25 is not supported because it can fail DNS resolution for app routes and external service calls on localhost.

1. Install dependencies:

• pnpm install

1. Copy the environment template:

• cp apps/web/.env.example apps/web/.env.local

1. Start a dedicated PostgreSQL instance for local development.
2. Apply the SQL migrations from packages/db/migrations.
3. Seed local development data when needed with packages/db/seed/0001_dev_seed.sql.
4. Configure Better Auth provider settings:

• Google OAuth if enabled
• email/password and verification settings as needed

1. Configure auth redirect URLs:

• Google OAuth must allow Better Auth's provider callback route:
• http://localhost:3000/api/auth/callback/google
• http://127.0.0.1:3000/api/auth/callback/google
• https://certscore.ai/api/auth/callback/google
• /auth/callback is only an app alias route; it is not the redirect URI Better Auth initiates with Google

1. Keep local and production auth isolated:

• local NEXT_PUBLIC_APP_URL should be http://localhost:3000
• local database and auth secrets should come from the dev environment
• production secrets should exist only in the active AWS/web or worker deployment settings

1. Create the S3-compatible bucket referenced by S3_BUCKET.
2. Install Playwright Chromium for the validation runtime:

• pnpm --filter @website-signal-risk-scanner/validation-worker exec playwright install chromium

1. Start validation local development with a watched validation worker:

• pnpm dev:validation

1. Start the main local app by itself when needed:

• pnpm --filter @website-signal-risk-scanner/web dev

1. Use WS01 when you need the standalone scanner locally.
2. Start the standalone scanner locally against the same dev database and storage env as localhost:3000:

• pnpm dev:scanner:local

1. Use the combined local runner only when you want web + validation together in WC01:

• pnpm dev:all

1. Run a validation scheduler sweep manually when needed:

• pnpm dev:validation:scheduler

Development verification

Use these commands before shipping changes:

• pnpm turbo run typecheck
• pnpm turbo run build

Validation-specific checks:

• pnpm --filter @website-signal-risk-scanner/validation-worker typecheck
• pnpm test:scan-pipeline

The scan pipeline test is deterministic and runs locally from apps/validation-worker/src/validation/pipeline.test.ts.

The normalized concern lifecycle in WC01 is documented in docs/normalized-concern-pipeline.md.

CI validation

GitHub Actions workflow: .github/workflows/accessibility-validation.yml

• worker-scan-pipeline-tests runs on pushes to main, pull requests, and manual dispatch. It installs Chromium, typechecks validation-worker, and runs pnpm test:scan-pipeline.
• live-validation-smoke runs after the deterministic job and executes pnpm --filter @website-signal-risk-scanner/validation-worker smoke:validation only when the runtime secrets are configured.
• If those secrets are missing, the live smoke job is skipped and only the deterministic scan pipeline job runs.

Runtime validation tooling

Use these lightweight checks before first deployment validation:

• pnpm dev:storage:local
• pnpm --filter @website-signal-risk-scanner/web check-env
• pnpm check-env:validation
• pnpm --filter @website-signal-risk-scanner/validation-worker check-env
• pnpm --filter @website-signal-risk-scanner/validation-worker check-runtime

For local validation runs, pnpm dev:storage:local starts MinIO against the apps/web/.env.local S3 settings and creates the configured bucket when needed. Run it alongside pnpm dev:scanner:local and pnpm dev:validation:worker.

Use this runtime smoke helper:

• pnpm --filter @website-signal-risk-scanner/validation-worker scheduler

The full runtime QA sequence is documented in docs/runtime-validation.md. The validation pipeline design and deployment shape are documented in docs/validation-pipeline-plan.md. The validation crawler deployment and VM runbook is documented in docs/validation-ops-runbook.md. Cloudflare Verified Bot setup is documented in docs/cloudflare-web-bot-auth.md.

Web Bot Auth

ConsentCheck can expose a signed HTTP Message Signatures key directory and sign outbound HTTP crawler requests for Cloudflare Verified Bot workflows.

Required configuration:

• WEB_BOT_AUTH_ENABLED
• WEB_BOT_AUTH_PRIVATE_KEY_PEM
• WEB_BOT_AUTH_SIGNATURE_AGENT_URL
• WEB_BOT_AUTH_EXPIRES_SECONDS
• WEB_BOT_AUTH_INCLUDE_NONCE

Common commands:

• pnpm web-bot-auth:generate
• pnpm web-bot-auth:print
• pnpm web-bot-auth:test-request
• pnpm test:web-bot-auth

Production deployment

Web runtime

• the live public web deployment target is the AWS ECS/Fargate path for certscore.ai
• consentcheck.site is owned outside WC01 and must not be deployed by this repo
• main deploys through .github/workflows/web-aws-ecs-deploy.yml
• run pnpm ops:check:deploy before or after topology changes to catch stale local assumptions
• run pnpm ops:check:live against the public hosts to verify runtime target and revision alignment after deploy
• treat docs/aws-web-ecs-cutover-plan.md as the active web deployment runbook
• treat docs/deploy-amplify.md as future-state reference material only

Validation And Scanner Runtime

• do not use WC01 for the primary scanner deploy path
• use WS01 for scanner runtime deployment
• keep WC01 deployment guidance scoped to web and validation only
• keep web production, validation worker, and scanner runtime as three distinct deployment paths
• use the AWS validation deployment lane for WC01 validation runtime changes

Database, Auth, and Storage

• create production PostgreSQL, Better Auth, and S3-compatible storage resources separate from local development
• apply all migrations in order
• configure the production site URL and production redirect URLs only
• create the report storage bucket referenced by S3_BUCKET

Scheduler

Recommended production trigger:

• run the validation scheduler sweep every hour if the validation runtime is enabled

Command:

• pnpm dev:validation:scheduler

The primary scanner scheduler now lives in WS01.

Deployment readiness checklist

• Environment variables configured
• PostgreSQL database created
• database migrations applied
• Better Auth providers configured
• S3-compatible storage bucket created
• Playwright browsers installed
• Worker process running
• Scheduler cron configured
• First domain scanned successfully
• Report generated successfully
• PDF export generated successfully

First production validation

After deployment, validate in this order:

1. run the env and runtime checks above
2. verify login and organization bootstrap
3. verify preview scan flow
4. verify full scan completion
5. verify findings, scores, and report persistence
6. verify PDF export
7. verify regression on a second scan
8. verify the hourly scheduler sweep path

Operational notes

• The standalone scanner in WS01 owns crawl, auditing, scoring, reporting, PDF generation, and schedu