Changelog
Every notable change to LaunchKit AI, newest first.
1.1.1
Changed
- The licence states lifetime updates and support on every tier, names the seats each tier includes (1, 3 and 10 developers), sets a 30-day no-questions refund, and gives a real contact address. The template warning is gone.
SECURITY.mdpoints at a monitored address instead of a placeholder.
1.1.0
Added
- Google sign-in: a Continue with Google button on the sign-in and sign-up pages, shown once
GOOGLE_CLIENT_IDandGOOGLE_CLIENT_SECRETare set. A first sign-in creates the workspace and its credits like any other sign-up, and an existing password account is linked when Google reports the same verified address.
1.0.2
Fixed
- CI sets the session secret the app requires, so the unit tests, build, evals and browser tests all run in GitHub Actions. A test now fails if the workflow ever misses a required environment variable.
1.0.1
Fixed
- Long conversation titles no longer run underneath the pin, rename and delete buttons.
- The design gallery browser test passes against a production build, where the gallery is hidden, so CI is green again.
- The landing page FAQ now matches the licence: client work is allowed on every tier.
1.0.0
The first release.
Added
- Foundation: Next.js 16 with TypeScript strict, Biome, Tailwind CSS v4 and a token-based design system with contrast tests; Zod-validated environment; PostgreSQL through Prisma 7; GitHub Actions CI.
- Auth: email and password with verification, password reset by email, two-factor authentication with an authenticator app and backup codes, an account menu, and a welcome email the first time someone confirms their address.
- Workspaces and teams: a personal workspace on sign-up, a workspace switcher, invitations that bring people back after they sign in or up, owner/admin/member roles, seat limits and an activity log.
- Billing: plans in
src/config/plans.ts, Stripe Checkout and the billing portal, idempotent webhooks, plan credits topped up on upgrade, one-off credit packs, and owner emails and in-app notices when a plan starts, a trial is ending or a payment fails. - AI credits: an append-only ledger with reserve-and-settle, monthly grants, low-credit emails and in-app notices, a banner on every workspace page when credits run low, monthly spending caps with a warning for owners, and a usage page with daily spend and a breakdown by feature.
- Chat: streaming replies as markdown with syntax-highlighted code, copy buttons, stop generating (billed only for what streamed), titles written by the fast model, history grouped by date with rename and delete, regenerate and edit-and-resend that never lose the old reply on failure, pinning, search across titles and messages, Markdown export, read-only workspace tools the model can call (credit balance, recent conversations, usage this month), and conversation summaries.
- Public API: scoped API keys with a rate limit per key, a cursor-paginated conversations endpoint, an OpenAPI 3.1 description, and signed outbound webhooks with a delivery log and redelivery.
- Account data: a JSON export of everything the app holds about a person, and self-service account deletion, confirmed by email, that never strands a team or leaves a paid plan billing.
- Notifications: an in-app bell and per-person notification settings.
- Onboarding: a getting-started checklist and removable sample data.
- Internationalisation: next-intl with English and a Spanish demo.
- Admin: a platform dashboard with revenue, margin, churn and per-workspace AI economics, workspace search, credit adjustments with a recorded reason, feature flags, and a kill switch that pauses all AI without a redeploy.
- App shell: a ⌘K command palette, a mobile menu, and a development-only gallery of every component at
/design. - Operations: a public
/api/healthendpoint for uptime monitors that never reveals why a check failed, and structured logs. - Marketing: landing, pricing, changelog and legal template pages with sitemap and robots routes, link preview images and structured data for search engines.
- Evals: a golden-dataset runner with rule and judge-model graders, a saved baseline that fails CI on a regression, and cost reporting.
- Developer experience:
bun run setup, agent instructions, task recipes, module and public API endpoint generators, a seed script, a Stripe bootstrap script, a Docker Compose file for the database, and CI checks that fail on schema drift or falling test coverage.
Security
- Content moderation before a chat message spends credits, optional redaction of personal data before it reaches the model, and a system prompt hardened against prompt injection.
- A nonce-based Content-Security-Policy and security headers.
- Rate limits on chat, the API, invitations, password resets, confirmation emails and data exports.
- Sign-in locks for 15 minutes after five wrong passwords for one address, without revealing whether the account exists.
- Email confirmation before joining a workspace.
bun auditin CI, a security policy and a pre-launch security checklist.
Fixed
- Plan credits now top up when a subscription changes.
- A paying workspace can no longer start a second subscription.
- Customers who run out of credits are told why and linked to plans.
- Invitations are rate limited per person.
- Listings with identical timestamps are ordered deterministically.
- A new API key is shown once without writing cookies during render.
- A chat turn the workspace cannot pay for no longer leaves an empty conversation or an unanswered message.
- The monthly credit allowance renews the first time a workspace spends or views its balance in a new month.
Docs
- A docs index, architecture overview, configuration reference, deployment guide, models and costs guide, troubleshooting, upgrade guide and feature guides, several of them checked against the code by tests.