Developer resources
Free guides for building AI-native products.
A practical reference for developers, drawn from real production work. No sign up, no paywall. Pick the tools you need, then follow the operational guides that decide whether a product holds up once it is live.
Where to start
Choose your layers
Decide which parts of the stack your product actually needs. Most products touch edge, data, identity, and one or two others.
Read the integration references
Each technology below has a short note on what it is good for, so you can pick the right tool without trawling ten docs sites.
Follow the operational guides
Setup, secrets, security, deployment, and testing. The unglamorous parts that decide whether a product holds up in production.
Ship, then instrument
Get it live, then make failures visible. You cannot fix what you cannot see.
Operational guides
The parts that decide reliability
Project setup
Spin up a new project: repository, framework, and a first deploy you can build on. Get to a live URL on day one.
Environment and secrets
Keep secrets server side and out of the client bundle. Centralize them so a key rotation is one change, not ten.
Security baseline
Validate every input at the boundary, verify all webhooks by signature, and set protective response headers from the start.
Deployment
Push to deploy, a preview for every branch, and a clean way to roll forward when something slips through.
Testing when you ship fast
A small team cannot test everything. Here is what to cover first so speed does not cost you reliability.
Vendor neutral AI
Route one code interface across several model providers, with fallback and cost visibility, so provider choice is config not code.
Prompt patterns that hold up
Reusable prompt structures for structured output, grounding, and refusal, written so they survive a model upgrade.
Mobile money done right
Take M-Pesa as a first class rail: phone normalization, integer money, token refresh, and idempotent callbacks.
Designing for emerging markets
Low bandwidth, feature phones, and patchy connectivity. How to build for the device people actually carry.
Data privacy and residency
An honest posture for regulated data: where it lives, who can touch it, and how to prove it when a buyer asks.
Integration reference
Thirty five technologies, by layer
A one line note on what each tool is good for, so you can choose the right one without reading ten documentation sites first.
GitHub
Source of truth and deploy trigger; push to ship. Drive PRs and Actions via CLI.
Infisical
Centralized secrets via machine identity; open-source and self-hostable for data residency.
Visual Studio
Enterprise and .NET IDE; out-of-process extensibility SDK for tooling.
Chrome DevTools
Real-browser automation plus performance and UI verification.
Context7
Live, version-accurate library docs to ground AI-assisted code (vendor-neutral).
Microsoft Learn (MCP)
Grounds AI in current Azure, .NET, and Entra docs; research engine for Microsoft builds.
Vercel
Primary push-to-deploy host; per-PR previews; edge and serverless.
Cloudflare
Workers, D1, KV, and R2 (zero-egress media), the global edge.
Netlify
Parallel git-driven host; Forms and Identity when needed.
Render
Always-on services, workers, cron, and persistent connections.
Google Cloud
Heavyweight: Compute, Cloud Run, BigQuery, Vertex AI; IAM unifier.
Supabase
Postgres plus Auth, Storage, and row-level security; ships pgvector.
Neon
Serverless Postgres with instant per-branch databases; scales to zero.
Upstash
HTTP Redis and QStash queues that run on the edge; rate-limiting and jobs.
Pinecone
Managed vector DB with integrated inference; the retrieval layer.
pgvector
Embeddings inside Postgres; joinable with business data, the cheapest RAG.
Vercel AI SDK + Gateway
The unification layer: one interface, multi-provider, fallback, cost dashboards.
OpenAI
Structured output and function-calling strength.
Google AI Studio (Gemini)
Multimodal and image (powers the thumbnail pipeline); Developer-API path.
DeepSeek
Budget reasoning tier; OpenAI-compatible drop-in for cost-sensitive volume.
Hugging Face
Open-source, self-hosted, and fine-tuned models (e.g. Swahili-tuned).
ElevenLabs
TTS and STT, multilingual voice (IVR, voice notes, Swahili and Sheng).
Anthropic
One reasoning and codegen provider within the vendor-neutral routing.
Clerk
Managed auth: sign-in, sessions, organizations; webhook verification.
M-Pesa / Daraja
First-class mobile-money rail; idempotent callbacks, phone and amount rules.
Stripe
International cards and subscriptions; complements M-Pesa.
Africa's Talking
SMS, USSD (feature phones, no data), and voice, reach everyone.
WhatsApp Business API
Dominant chat channel; 24-hour window plus template messages.
Resend
Transactional email (React Email); secondary to SMS and WhatsApp in-market.
Figma
Design-to-code in both directions; grounds UI work in real designs.
Canva
Data-driven, on-brand graphic mass-production (Brand Templates and autofill).
Notion
Docs and PM hub; spec-to-ticket, knowledge capture.
Sentry
Error and issue tracking; surfaces production failures with stack traces.
Porkbun
Registrar plus programmatic DNS (wires the codeamanilabs.org subdomains).
Domain Portfolio
Strategy layer above the registrar: which names to build, in what order.
The knowledge hub goes layer by layer with full guides, and the engineering blog shares field notes from production.
Building something with this stack?
If you would rather have it built and run for you, tell us what you have in mind.