Security
On-prem by design
Your tokens, your network, your audit trail. Issue text never leaves your infrastructure for AI analysis.
Where does data live?
Postgres on infrastructure you operate. Issue metadata synced from VCS; PATs encrypted when TOKEN_ENCRYPTION_KEY is set.
Does issue text go to vendor AI?
No. LLM analysis uses local Ollama on your network. Workers read Postgres only for analysis jobs.
How do users sign in?
GitHub/GitLab OAuth. Production requires email allowlist or pre-provisioned invites.
Who can access the app?
Closed registration after setup — admins invite emails + roles before first login.
How are VCS tokens stored?
AES-256-GCM when encryption key configured; fail-fast in production if missing.
Internet exposure?
You control network placement. Bundle does not expose Postgres/Redis on host ports. HTTPS via your reverse proxy.
Rate limiting?
Redis-backed limits on API routes; configurable per route type via RATE_LIMIT_* env vars.
Top reviewer questions
Is this multi-tenant SaaS?
No — single-tenant instance per install.
Can any GitHub user sign in?
Not in production — allowlist + invite model.
What scopes for PATs?
Read for sync; write scopes only if you enable write-back apply.
SSO / SAML?
OAuth via GitHub/GitLab today; direct enterprise SSO is roadmap, not v1.
Audit trail?
Admin audit log for admin actions, sync, analyze, apply, invites.