App Development Cost 2026: Real Numbers for 20+ App Types
Last updated: 14 April 2026Region: Global (USD)Data source: MyAppTemplates shipped-apps corpus + public SOW benchmarks
Executive Summary
The median agency quote to build a production mobile app in 2026 sits at $50,000–$115,000, with 3–5-month timelines. Independent builders running Claude Code on top of a production boilerplate ship the same app for $199 boilerplate plus $30–$300 in API spend — total well under $500 for most apps, shipped in days to two weeks instead of months.
The unlock isn't AI, it's AI with a boilerplate. Builders running agentic tools from scratch — no scaffolding — burn through $500–$1,500 in tokens and 5–8 weeks rebuilding auth, billing, CI, and schema that has already been written ten thousand times. That's the baseline setup spend a production boilerplate removes, and it's the comparison that matters if your real choice is scratch versus starting from a working foundation.
Apps with the cleanest DIY savings: SaaS dashboards, fitness trackers, and on-demand marketplaces. Categories where an agency engagement is usually the right call: HIPAA-regulated health, regulated fintech, and licensed media (streaming, audio) — the hard costs in those categories are compliance, audit, and content licensing, and the boilerplate path doesn't replace specialist delivery.
The DIY path is not a like-for-like swap for an agency relationship. It assumes a hands-on operator who owns the build daily, reviews migrations before shipping, and is comfortable making architectural calls. Agencies remain the right choice when you need outsourced execution, named delivery responsibility, or procurement cover.
Data
App Development Cost by App Type (2026)
Median agency quote vs. DIY (boilerplate + AI tokens), ranked by savings.
Every DIY build starts with the same flat boilerplate fee:$199 one-time — column below shows marginal Claude Code API spend on top
Marketplace and on-demand apps produce the biggest absolute savings — an Uber-style app is $100k–$220k at an agency across 3–5 months versus roughly $419 DIY ($199 boilerplate + ~$220 in Claude Code API spend) across 1–2 weeks. 70%+ of the code is standard scaffolding: auth, modular architecture, edge runtime, CI, Sentry, and a working billing abstraction. The boilerplate ships that foundation — and the Kilo subagents to let Claude Code build on top of it — so you don't spend week one on auth and week two on CI.
Agency Quote$100,000 – $220,00012–20 weeks, 3-person team
DIY Cost$199 boilerplate + ~$220 AI= ~$419 total, 1–2 weeks solo running Claude Code with subagents
What the boilerplate actually gives youPre-wired: JWT + phone-OTP auth for both rider and driver apps, Cloudflare Workers + D1 + Drizzle runtime, billing abstraction layer, Sentry, GitHub Actions CI, and the Kilo subagents (@backend-dev, @mobile-dev) that make Claude Code productive from the first prompt. You still build Stripe Connect wiring, real-time tracking channels, push, and KYC — but against working foundation, with Claude Code doing roughly a day's work per feature.
Still hire an agency ifYou need >10,000 concurrent trips on day one, or you're a funded startup where a 10-week delay costs more than the $120k quote.
2. SaaS & Productivity (Fastest ship time)
SaaS dashboards ship in 3–5 days because CRUD + auth + billing covers 80% of the scope. Agency quotes in this category typically run $55k–$110k across 8–12 weeks — the right call when you need outsourced delivery, security review, or procurement cover. For hands-on founders shipping their own product, the boilerplate covers most of the foundation work that drives the bulk of those hours.
Spotlight Build
SaaS Dashboard (B2B)
The BuildEmail + magic-link auth, team workspaces, Stripe subscriptions with entitlements, admin panel, onboarding, transactional email.
Agency Quote$55,000 – $110,0008–12 weeks
DIY Cost$199 boilerplate + ~$55 AI= ~$254 total, 3–5 days solo with Claude Code
Where the agency hours goA typical $55k–$110k SOW spends ~200 hours on Stripe billing plumbing, role-based access control, audit logging, and CI/CD. Both billing and RBAC are boilerplate defaults, so on the DIY path the AI spend goes into product surface — screen layout, brand, and the workflow that's actually unique to your app — rather than into rebuilding billing infrastructure.
When an agency is still the right callIf you need SOC 2 evidence on day one, named delivery responsibility for procurement, or a team to own the build while you focus elsewhere, an agency engagement is doing different work than the boilerplate is.
How to ship at these numbers
The DIY column isn't magic — it's a 3-step process. Skipping step 1 is the most common reason token bills come in 5–10× higher than expected.
1
Scope before you write a single prompt
Lock your feature list in a written spec before spending a token. AI without architectural discipline burns tokens rebuilding the same scaffolding repeatedly — that's the largest avoidable line item in the 2026 data.
2
Start with a boilerplate, not a blank editor
Every minute the model spends rebuilding auth, session handling, Stripe webhooks, or CI is minutes you're paying to recreate code that's already been written ten thousand times.
3
AI pair-coding, not AI vibe-coding
Give the model one bounded task at a time, with explicit acceptance criteria. Review the diff. Never let it "figure out the architecture."
Frequently Asked Questions
Why is the AI spend so low given the complexity?
Once the boilerplate has pre-wired auth, billing, CI, data layer, and edge runtime, Claude Code's job shrinks to building the feature that makes your app different. For a solo developer running agentic tools (Claude Code + Kilo subagents + slash commands), that's 8–12M tokens across 1–2 weeks for a complex build — an Opus/Sonnet blend lands around $180–$280 in API. Simple apps are 2–4M tokens, $30–$60 in API. The boilerplate's $199 is still the dominant line item for most builds.
Why is the agency quote range so wide?
Agency quotes vary 2–3× based on region (London vs. Eastern Europe), team seniority, and how much the agency thinks they can charge given your funding profile. The low end is typically an Eastern European shop; the high end is a London or SF boutique quoting a VC-funded startup.
How is this different from low-code tools like Bubble or Adalo?
Different category. Low-code platforms keep your app inside their runtime, which is a reasonable trade for non-technical founders who want speed and don't plan to leave the platform. A boilerplate gives you real code on standard infrastructure (React Native, Hono, Cloudflare), so you don't rewrite at scale and you're insulated from platform pricing changes. Pick the one that matches how technical your team is and how portable you need the codebase to be.
What if my app needs custom native modules?
Add $30–$150 in AI spend per major native integration (Bluetooth LE, advanced camera, CoreML). React Native handles most of it via Expo modules. Genuinely novel native work — bespoke hardware integrations, low-latency audio/video, advanced ML on-device — is exactly where agency teams add real value.
Why are Spotify and Netflix marked "license-gated"?
The software for a streaming app is ~$275–$340 in Claude Code. The content licensing (music publishers, film distributors) is $600,000+ and has nothing to do with code. The boilerplate can't help you there, and neither can an agency for that matter.
What about the $600 AI spend for banking?
That's the software build cost, and it's higher than other categories because of the volume of regulation-driven edge cases the AI has to handle. Banking also requires FCA authorisation (UK) or FinCEN registration (US) — a $30k+ compliance exercise with specialist lawyers. Neither the boilerplate nor an agency's quote usually includes that; it's a parallel workstream.
Software-first app costs have shifted.
For many software-first apps in 2026, a large share of the initial build cost is now foundation work rather than novel product logic. Starting from $199 boilerplate plus a few hundred dollars of AI spend can materially shrink that setup phase for hands-on teams, while agencies still make sense when you need outsourced delivery or compliance-heavy execution.