51 terms, translated twice.

Plain English

A set of functions your app exposes so other apps can talk to it.

MGT-Speak

If you want your website to send data to Zapier, your CRM, or a mobile app, you need an API. Usually already included.

Plain English

A network of servers worldwide that cache your site near your visitors.

MGT-Speak

Why your site loads in 200ms instead of 2 seconds when someone visits from Australia. Cloudflare + Vercel handle this automatically.

Plain English

Browser security rule that blocks one site from calling another site's API without permission.

MGT-Speak

The reason your widget embed throws errors in the console. Fixable in 5 min if the backend is yours, painful if it's someone else's.

Plain English

Create, Read, Update, Delete. The four things an app does with data.

MGT-Speak

Every internal tool is 80% CRUD. When we estimate a dashboard, we're really estimating 20-40 CRUD operations against a schema.

Plain English

The styling language that controls how a website looks.

MGT-Speak

Tailwind CSS is our default because it keeps design tokens in the markup. Consistency across pages without a design system babysitter.

Plain English

The system that turns a domain name into the server IP your browser connects to.

MGT-Speak

If your site 'works' at vercel.app but not at yourdomain.com, it's DNS. Cloudflare has a 5-min fix 98% of the time.

Plain English

A set of pre-built tools for building apps of a specific shape.

MGT-Speak

We default to Next.js for web, FastAPI for Python backends. Picking a framework is the single biggest dev-time lever in a project.

Plain English

Code that runs between a request hitting your server and your actual logic.

MGT-Speak

Auth checks, rate limits, i18n routing, A/B test assignment. On Vercel this runs in Fluid Compute at the edge.

Plain English

Single-Page App. Loads one HTML shell, then JavaScript handles all navigation.

MGT-Speak

Great for logged-in dashboards, bad for public SEO pages. We use hybrid rendering (Next.js App Router) so marketing is crawlable but apps feel instant.

Plain English

A URL that another service hits when something happens.

MGT-Speak

Stripe calls your webhook when a payment succeeds. Discord calls yours when a message is posted. The glue between all the SaaS tools in your stack.

Plain English

Object-Relational Mapper. Translates database rows into code objects.

MGT-Speak

Prisma is our default. Saves 30-40% of backend dev time and catches schema mismatches at compile time instead of 3am in production.

Plain English

Query language for APIs. One endpoint, you specify exactly what data you want.

MGT-Speak

Great for mobile apps that need varying data shapes. Overkill for most web dashboards. REST + tRPC usually wins for our scope.

Plain English

The most common API style. URLs map to resources, verbs (GET/POST) map to actions.

MGT-Speak

Every public SaaS API is REST. Boring, well-understood, trivial to document. Default choice unless you have a specific reason otherwise.

Plain English

The structure of your data: what tables exist, what columns, what types.

MGT-Speak

We sketch this first on any project because schema changes post-launch are where 50% of bugs come from. Get it right in week 1.

Plain English

A versioned change to your database schema.

MGT-Speak

Every prod schema change is a migration. Bad migrations = downtime. We never run `db push` in production.

Plain English

Anti-pattern where code loops and makes one database call per iteration.

MGT-Speak

The #1 cause of 'suddenly slow' dashboards. Usually a 5-line fix, but only if you know to look.

Plain English

Storing a result so the next request doesn't have to recompute it.

MGT-Speak

Next.js App Router caches aggressively by default. Most perf wins on existing sites are cache-config fixes, not code rewrites.

Plain English

Standard for 'Sign in with X'. You prove who you are via Google/GitHub/Discord without giving us your password.

MGT-Speak

Discord OAuth is why Service Plug + Holy Services can trust gamertags. Every B2B app should offer Google OAuth from day one.

Plain English

JSON Web Token. A signed string that proves who you are between requests.

MGT-Speak

Session plumbing. Fine for most apps. Better Auth handles rotation + revocation so you don't have to.

Plain English

Two-factor / multi-factor authentication. You prove identity with something you know + something you have.

MGT-Speak

Non-optional for admin panels. We enforce it on all /admin routes by default. Authy and TOTP apps are the safe-bet choice.

Plain English

Role-Based Access Control. Permissions tied to roles (admin, editor, viewer).

MGT-Speak

LaunchWise has 21 permissions across 7 roles. Any B2B app with more than 3 users needs this from day one.

Plain English

Cross-Site Request Forgery. Attack where a malicious site tricks your browser into making authenticated requests.

MGT-Speak

Next.js App Router has built-in protection. Still need to audit form actions. Fresh-reviewed on every project.

Plain English

Cross-Site Scripting. Attacker injects JavaScript into your page that runs in other users' browsers.

MGT-Speak

React auto-escapes by default, but user-generated HTML (rich text, markdown) is where it bites. We never innerHTML raw content without a sanitizer.

Plain English

The encryption that makes URLs start with https:// instead of http://.

MGT-Speak

Let's Encrypt is free and automatic on Vercel / Cloudflare. If your site doesn't have it in 2026, that's a red flag before any other concern.

Plain English

Limiting how many requests a user can make in a time window.

MGT-Speak

Required on contact forms, login, any public API. Without it, one bot can rack up $500 in Stripe API fees overnight.

Plain English

Server-Side Rendering. HTML is built on the server each request.

MGT-Speak

Default for logged-in pages in Next.js. Best of both worlds: fast like static, dynamic like an app.

Plain English

Static Site Generation. HTML is built once at build time and served from a CDN.

MGT-Speak

Marketing pages, blog posts, docs. Cheapest possible hosting, fastest possible loads. If content doesn't change per user, SSG it.

Plain English

Incremental Static Regeneration. Static page that rebuilds itself on a schedule or on-demand.

MGT-Speak

Ship-log. Ticker widgets. Anything that's 'static-ish' but should refresh every hour. Cheap and fast.

Plain English

Server code that runs in 300+ cities worldwide, close to users.

MGT-Speak

On Vercel, Fluid Compute covers most of this now. Use cases: A/B tests, geo-based redirects, auth guards before rendering.

Plain English

Continuous Integration / Deployment. Code gets tested + deployed automatically on push.

MGT-Speak

Vercel handles CD out of the box. We set up GitHub Actions for CI (tests + typecheck) on any project that grows past a single dev.

Plain English

A way to package an app with everything it needs to run, anywhere.

MGT-Speak

Required when deploying to Railway, Hetzner, or a client's own servers. Overkill for Vercel-hosted Next.js apps.

Plain English

Run two versions of an app side-by-side, flip traffic from old to new.

MGT-Speak

Vercel does this automatically with preview deployments + instant rollback. One of the main reasons we default to it.

Plain English

Instantly reverting to a previous working version when something breaks.

MGT-Speak

Vercel rollback is one click, 2 seconds. Make sure your hosting has this before you trust it with production.

Plain English

Percentage of time your service is available.

MGT-Speak

99.9% = 8.7 hours of downtime per year. 99.99% = 52 min. Most SaaS tools quietly sit at 99.5% or worse. Ask for the number.

Plain English

Large Language Model. The general class Claude, GPT, Gemini belong to.

MGT-Speak

Default provider choice matters more than people admit. Claude Sonnet is our default for production features because of tool-use reliability.

Plain English

Letting an LLM call your own functions as part of its response.

MGT-Speak

The reason LaunchWise AI assistant can actually create tasks, send emails, query the CRM. Without tool use, AI is just a chat toy.

Plain English

Retrieval-Augmented Generation. Pull relevant docs first, then ask the AI.

MGT-Speak

How you make an AI answer correctly about YOUR product. Vector DB + embeddings + a retrieval step. Not always needed, but often.

Plain English

A database that finds items by semantic meaning instead of exact match.

MGT-Speak

pgvector on Postgres covers 90% of use cases. Dedicated stores (Pinecone, Weaviate) only when you're past 100k vectors with heavy QPS.

Plain English

Storing parts of a prompt so repeat calls are cheaper + faster.

MGT-Speak

On Anthropic, 5-min TTL on cached blocks. Can cut token costs by 60-80% on any app with a stable system prompt.

Plain English

Training an AI on your specific data to behave a specific way.

MGT-Speak

Rarely worth it for most use cases. Good prompts + RAG + tool use beats fine-tuning in 90% of scenarios we've tested.

Plain English

Model Context Protocol. Standard for connecting AI agents to external tools.

MGT-Speak

The reason Claude Code can edit files, run commands, and talk to your infra. We build custom MCP servers for clients who want internal tooling.

Plain English

Minimum Viable Product. The smallest thing that delivers real value.

MGT-Speak

MVP scoping is where projects fail. We push hard for a 2-week MVP before anyone argues about V2 features. Most 'MVPs' are already V1.5.

Plain English

When a project keeps adding features beyond the original agreement.

MGT-Speak

Why fixed-price projects go over. We solve it with a written scope doc + change-order forms. Saying yes to one-off requests is how teams burn out.

Plain English

Short-term code choices that make future changes harder.

MGT-Speak

Every project has it. Danger is invisible debt: no tests, no docs, one person knows how it works. We audit this on every rescue engagement.

Plain English

Service Level Agreement. Written commitment on response time + uptime.

MGT-Speak

Maintenance plans have them. One-off projects don't. Ask for one in writing if your tool breaks and your business stops.

Plain English

Non-Disclosure Agreement. Both parties agree not to share confidential info.

MGT-Speak

Standard for any engagement touching user data, internal ops, or pre-launch product. We send a mutual one on every scoped project by default.

Plain English

Security compliance standard that enterprise buyers often require.

MGT-Speak

Expensive. ~$20-40K/year minimum. Not needed until you're selling to companies that ask for it. Don't volunteer it.

Plain English

EU data-privacy law. Applies if you have EU users, regardless of where you're based.

MGT-Speak

Cookie banners, right-to-delete endpoints, data-export flows. Baseline for any B2C site with >1k monthly visitors.

Plain English

Terms of Service. The contract users agree to by using your site.

MGT-Speak

Templates from Termly or Termageddon cover 80%. Custom review by a lawyer before any real money flows. Not legal advice.

Plain English

Transferring code, credentials, and docs to a new dev team.

MGT-Speak

Clean handoffs take 4-8 hours. Dirty handoffs take 4-8 weeks. We deliver a handoff packet on every finished engagement, unsolicited.

Plain English

Number of people who can be hit by a bus before your project breaks.

MGT-Speak

Solo-dev projects have a bus factor of 1. We reduce that with docs, tests, and readable code so any dev can pick it up.