Roll out LGTM across many repos

GitHub App installs at the ORG level (or personal-account level). A single install grants LGTM access to the repositories the installer selects — could be 1, could be 'all repos in org'.

Once installed, LGTM picks up webhook deliveries for all enrolled repos. The first PR opened on each enrolled repo triggers index + first review. No per-repo manual setup required.

Adding repos later: github.com/organizations/{org}/settings/installations → LGTM → Configure → add repos. New repos start getting reviews on the next PR opened against them.

Default review behavior is configured once at the org level: which agents enabled, which models, what verdict threshold, what severity blocks merge. New repos inherit these defaults.

Override per repo when needed: Settings → Repos → {repo} → Config. Override specific values; unset values inherit from org. Visual diff in the UI shows what's inherited vs overridden, so config drift is easy to detect.

Common pattern: org defaults are 'flagship model + strict severity + all agents'. Marketing/internal repos override to cheap model + relaxed severity. Production-critical repos override to flagship model + auto-approve disabled (human always reviews).

Don't enroll all 50 repos on day one. Phased rollout reduces noise + lets the team adjust thresholds before scale.

Week 1: enroll 1-3 'pilot' repos. Active monitoring of review quality, false-positive rate, team feedback. Adjust verdict threshold + agent selection based on what fires.

Week 2-3: expand to 10 repos covering different stacks (front-end / back-end / infra / etc.). Confirm cross-stack quality. Per-repo config tweaks for any with unusual patterns.

Week 4+: open enrollment. New repos opt in via a CODEOWNERS file or simple PR adding the repo to LGTM's selected list. Engineering team's responsibility, not central platform.

LGTM's pricing is per-account, not per-repo. Go ₹999 / Pro ₹1,999 / Max ₹9,999 per month covers 100 / 200 / 1,200 reviews respectively across all repos in your install (up to 5 / 20 / unlimited repos). No per-repo tax, no per-PR tax (just BYOK token cost which is a separate provider bill — or opt into Managed AI). Enterprise unlocks unlimited.

Cost scales with REVIEW VOLUME, not repo count. 100 repos with low PR activity might cost less than 10 repos with constant PR churn — because the BYOK token bill scales with token usage.

Org-level usage dashboard: Settings → Usage. Shows reviews per repo, tokens per agent, average review latency, per-provider spend. Useful for spotting expensive repos that could move to cheaper models.