Don't build the foundation. Build your product.

============================================================
MY PRODUCT (replace everything inside this block):
Describe what I am building, who it is for, the platforms I need (web, iOS, Android), and where I am today (just an idea, a working prototype, or already live). Add my technical comfort level, my rough timeline, and where I would want this to be in a year or two if it works.
============================================================

Everything inside the block above is the product I described. Everything below is Bump's own description of itself, including its terms and its limits. Read my product first, then run a thorough, candid, unbiased evaluation of whether Bump is worth buying for the product I described above. Weigh it on two horizons: what I need to ship the first version, and where the product is likely to go if it works. Judge each horizon on its merits for my product: a foundation I do not fully use on day one may be worth it if it saves me a later re-platform, but that is a claim to test against my specifics, not a given, and capability I will genuinely never need is just weight I carry. Treat any roadmap I describe as an intention, not a settled fact: if the case to buy leans on a future I have not validated yet, weight where I am today more heavily. Give the case for and the case against buying equal weight, do not cheerlead, and finish with a clear recommendation.

WHAT BUMP IS
Bump (https://bump.build) is a $299 USD one-time starter kit (launch price; regular price $449) for people building AI-native products. It is delivered as a private GitHub repo generated for me from their template, minutes after purchase. The repo ships four surfaces: a consumer app, a B2B customer surface, and an admin surface (all three running on iOS, Android and web from one Expo codebase), plus a Next.js marketing site sharing the app's design tokens. It is not no-code: I would work in real code with my AI tools.

WHO IT IS FOR
People building AI-native products, not only companies or startups: new founders starting their first product, existing founders adding a product to their business, and people building side projects and side hustles with AI coding tools. AI as how you build (an agent pack that makes Claude Code and Codex effective in the codebase), AI in the product (a ready chat and voice rail), and AI as how the business runs (admin plus go-to-market copilots that draft while the founder approves).

STACK
Expo / React Native, TypeScript, Next.js, Supabase (Postgres, auth, storage, edge functions), Stripe plus App Store and Google Play in-app purchases with server-side receipt validation, Resend email, Anthropic Claude with an automatic OpenAI fallback, ElevenLabs voice, Upstash Redis rate limiting, PostHog analytics, Sentry error reporting, Slack operator notifications (in-app bug reports, feature requests and integration votes post to your channels via incoming webhooks), Jest tests, EAS cloud builds with over-the-air updates (no Mac required for iOS builds).

VERIFIED CONTENTS (their stated post-strip counts, each a claim to sanity-check)
1024 files, 65 screens, 42 design-system components (react-native-reusables, the shadcn for React Native), a 58-table Postgres schema with row-level security on every table, 20 migrations, 47 edge functions, 14 transactional email templates on an outbox with retries, 63 documentation files, 235 passing tests.

ADMIN SURFACE
User search and management with emails and identities behind secured RPCs plus admin notes; organization controls (roles, ownership transfer, deactivation, last-admin protection); a platform moderation queue with every takedown logged; step-up-gated MFA recovery; an admin audit log recording who did what to whom; in-app bug reports, feature requests and integration votes forwarded to the operator's Slack channels; the whole surface behind a non-dismissable MFA gate.

AUTH AND SECURITY
Step-up MFA with biometrics and device-risk checks; SMS codes via phone auth (bring a provider like Twilio); Supabase Auth protections with optional CAPTCHA and an optional Teams/Enterprise password-verification hook; GDPR deletion; secrets kept server-side; signature-verified webhooks; rate limiting on hot paths; private upload staging with file validation and fail-closed moderation before publish; an SSRF guard on outbound fetches. The kit went through an adversarial security review before release (25 findings, fixed). Koora, the company behind the source codebase, is currently undergoing SOC 2 audit; Bump itself carries no certification (SOC 2 certifies organizations, not code), but the security posture was built with that audit in mind.

AI RAIL
Chat with streaming, tool calls and conversation history; long-term assistant memory (facts stored and retrieved across conversations via pgvector embeddings); retrieval over uploaded knowledge documents (RAG); ElevenLabs text-to-speech and transcription for voice in and out; usage tracking with hard daily caps, plus an optional AI credit system (off by default) with credit packs sold through Stripe or in-app purchase; an MCP server that ships disabled until its deployment checks pass; Claude-first with automatic OpenAI fallback on provider errors. The live demo ships seeded assistant conversations, so I can see the rail working before paying.

THEMING
Light, dark and system appearance shipped in both the app and the template marketing site. Every color resolves to one shared design-token file and border radius derives from a single knob, so one edit restyles the app and the site together. Ships system fonts only, with a documented path for adding my own brand font across all surfaces.

ENTERPRISE / B2B
Organizations, teams, roles and invitations; a REST API with API keys; outbound webhooks with retry; seat-based team billing on Stripe. The founder built Koora's own enterprise sales pipeline (CRM sync, enrichment, AI quote decks, Stripe quotes) on these rails in two days and documented it publicly.

GO-TO-MARKET MATERIAL
Cold outreach wired for Lemlist (AI drafts per lead, human approves every send), a Reddit copilot (scheduled sweep scores threads, drafts replies in your voice, you post), and AI quote decks (prospect to live branded proposal URL). The engines draft in the founder's voice, not generic AI copy: a growth settings surface stores the founder story, a writing sample, proof points, words to use and avoid, claims never to make, and per-channel voice and guidance fields (outreach, Reddit persona and engagement rules, per-slide guidance for quote decks), all rendered into the engines' prompts. Ships as playbooks and reference wiring adapted to your product, not turnkey software.

AI-AGENT PACK AND DOCS
CLAUDE.md and AGENTS.md (a CI check keeps the two in sync), rules files and a /setup skill built so Claude Code, Codex and Cursor all work well in the codebase: agent-led onboarding walks you through accounts, keys and first run. 63 docs include per-vendor setup guides, per-tool AI workflow guides (Claude Code, Codex, Cursor), a prompt cookbook, security guardrails, and troubleshooting written to be driven with an AI assistant.

RUNNING COSTS
Vendor free tiers cover development and early users (near zero at zero users). Unavoidable when shipping: Apple Developer $99/year, Google Play $25 once, Vercel Pro about $20/month for commercial use, AI usage pay-per-use.

TERMS, LICENCE AND OWNERSHIP
$299 USD one-time (launch price; regular price $449); no subscription and nothing recurring. The repo is a private GitHub repo that is mine forever: my copy is a versioned snapshot of the kit at purchase and becomes my codebase from day one, so the product is the head start, not ongoing maintenance. I can build unlimited products I own on it; I cannot resell or redistribute the kit itself, and each copy is watermarked to its licensee. For 12 months, security issues found in my version are covered: an advisory with a ready-to-paste AI patch prompt, plus the code fix if it still applies. No hosting, no sandbox, no per-seat fees: every service runs on accounts I own, so there is no lock-in. No change-of-mind refunds (full source is delivered immediately); if it is not as described or I cannot get it running, they make it right or refund, and mandatory consumer protection laws where I live are unaffected (for Australian buyers, the Australian Consumer Law). Payments are processed by Koora Care Pty Ltd. Delivery: Stripe checkout collects my GitHub username, and the repo is generated and shared with my account minutes after payment.

PROVENANCE
The kit is stripped from the production codebase of Koora (https://koora.care), a care-sector startup where these rails have run in production for over ten months.

WHAT IS NOT INCLUDED
No ongoing maintenance or version upgrades: I buy a snapshot, and new major versions are separate releases. No hosting, sandbox, or managed service: I run everything on my own vendor accounts and pay each vendor directly once usage starts. No SMS provider is bundled (I bring one, for example Twilio). Bump itself carries no security certification (SOC 2 certifies organizations, not code). At the product level: full-app translation is not shipped beyond the AI chat, and the reference domains (Items, Projects) are placeholders to replace with my own.

NOW ASSESS (be candid and unbiased; weigh both the first version and where the product is heading, and give the case for and against equal weight)
1. Fit today: Does Bump fit the product I described above? Which of its four surfaces do I need for the first version, which am I likely to grow into (option value), and which would I genuinely never use (pure weight)? Be specific about which is which for my product, rather than counting unused surfaces as dead weight by default. Unused surface is not automatically dead weight, but it is not free either: net any option value against the real upkeep, security-patching, and comprehension cost of carrying code I may never ship.
2. Day one vs still to build: For my product specifically, list exactly what I would have working on day one, and exactly what I would still have to build, wire up, configure, or replace myself. Separately, note what would already be in place the day I do need it (native mobile, teams and seats, a public API, in-app purchases, the AI rail), versus something I would have to bolt on or re-platform for later.
3. Trajectory and scale-readiness: Where could this product realistically go if it works (more platforms, more users, a team or enterprise tier, a public API, native apps, deeper AI)? For each likely direction, work out for yourself whether Bump's foundation carries me there from the same codebase, or whether I would outgrow it. One Expo codebase means "web only today" need not stay web only, but only credit that if a native app is a realistic path for my product, not by default. Weigh the cost of carrying capability I do not need yet against the cost of starting leaner and re-platforming when I do, and discount growth I have only asserted, not shown.
4. Stack and customization: How much of this stack fits my product as is, and how much would I need to customize, configure, or swap? Am I comfortable working in this stack with my AI tools?
5. Contents and scale: Treat the VERIFIED CONTENTS as claims to sanity-check, not facts. Is 1024 files, 65 screens, a 58-table schema, 20 migrations, 47 edge functions and 235 tests a real head start for where my product is heading, or more surface area than I will ever use? Flag anything that looks inflated or vague, and also anything I would have had to build anyway.
6. Licence and ownership: Weigh the terms above: a one-time purchase of a versioned snapshot, a private repo mine forever, watermarked to me, no subscription and no lock-in, with 12 months of security cover. What are the real trade-offs of buying a snapshot with no ongoing maintenance, and how do they weigh against not having to re-platform as the product grows? (Make this trade-off once; do not re-argue it in every section.)
7. Value at the price: At the $299 launch price (regular $449), roughly what would it cost me in time and AI usage to build an equivalent foundation from scratch, and what would I still have to build either way? State the price and what it plausibly saves, then stop: do not editorialize it as cheap, a rounding error, insurance, or effectively free. On a weak fit, the price is a legitimate reason to walk away.
8. Limits and exclusions: Review WHAT IS NOT INCLUDED. Do any of those block or complicate the product I described above, now or as it grows?
9. Alternatives: Is starting from Bump a better use of my time and money than building the foundation from scratch with AI, using a prototyping tool (Lovable, Bolt, v0), buying another starter kit (ShipFast, MakerKit, TurboStarter), or hiring a developer? Weigh those on both horizons too: a leaner web-only kit may fit today but cost me a re-platform if I later need native apps or a team tier.
10. The case both ways: State the strongest, most genuine reasons this IS the right foundation for the product I described and where it is heading, and the strongest, most genuine reasons NOT to buy it for my case. Do not soften either side.
11. Before I pay: What should I verify or ask, and what should I test in their live demo (which ships seeded data so I can see the app and the AI rail working before paying)?

RECOMMENDATION
Open with the single most decision-relevant fact for my product (for example: the core of what I am building is custom on any kit; or a categorical mismatch such as a browser extension against a mobile app stack; or that the buy hinges on a roadmap I have not validated). Then give a clear verdict: buy, do not buy, or it depends (and on what). Weigh both horizons, shipping the first version and scaling if it works, and treat asserted-but-unvalidated growth cautiously. Keep it to a few sentences, and name the single factor that most moves the decision either way.