Building with AI
Go beyond autocomplete — connect tools, ship API integrations, and automate multi-step dev workflows.
This section is for developers who integrate AI: MCP servers that expose your data, LLM APIs in Next.js routes, agent loops in the IDE, and RAG when @Files is not enough. Each guide includes runnable patterns; tutorials walk through full projects.
How the Pages Relate
| Page | When you need it |
|---|---|
| Building MCP Servers | Connect DBs, APIs, or docs to Cursor/Claude Code |
| MCP Security | Auth, secrets, and team allowlists for MCP |
| Agent Skills | Portable SKILL.md workflows for repeatable agent tasks |
| LLM APIs | Ship streaming chat or tool calling in your app |
| Structured Outputs | Reliable JSON from LLMs with Zod schemas |
| AI Code Review | PR review workflows with human gates |
| AI in CI/CD | Review bots and failure triage in pipelines |
| Anthropic API | Claude models, extended thinking, prompt caching |
| OpenAI API | GPT models, structured outputs, vision, embeddings |
| Agentic Workflows | Plan → implement → review in the IDE or CI |
| RAG for Codebases | Large repos or product features over private docs |
| LLM Observability and Evals | Trace calls, monitor tokens, run quality evals |
Typical progression: MCP for personal coding → LLM APIs when shipping a feature → RAG when retrieval quality matters → Observability before launch.
Learning Paths
Solo developer
- Building MCP Servers — extend your agent locally
- MCP Server Tutorial — build and test end to end
- LLM APIs — streaming and tool schemas
- Anthropic API or OpenAI API — provider-specific setup
- Streaming Chat Tutorial — full Next.js chat app
- Agentic Workflows — safe multi-step automation
Team lead
- Building MCP Servers — approved servers and secrets policy
- RAG for Codebases — data residency and what not to index
- Security and Prompt Injection — production LLM features
- LLM Observability and Evals — monitoring before launch
- Team AI Policy — CI gates for agent automation
Start Here If…
| Your goal | Start with |
|---|---|
| Connect Postgres or internal API to Cursor | Building MCP Servers |
| First MCP server from scratch | MCP Server Tutorial |
| Add chat to your Next.js app | Streaming Chat Tutorial |
| Integrate Anthropic / Claude directly | Anthropic API |
| Integrate OpenAI / GPT directly | OpenAI API |
| Model misses docs in a large monorepo | RAG for Codebases |
| Agent keeps over-scoping refactors | Agentic Workflows |
| Production monitoring and quality evals | LLM Observability and Evals |
| Compare Bedrock vs direct API | AI Platforms |
Guides
Building MCP Servers
Expose databases, APIs, and docs to your agent.
Agent Skills
Author SKILL.md packages for repeatable agent workflows.
LLM APIs
OpenAI, Anthropic, Vercel AI SDK, streaming, tool calling.
Anthropic API
Claude models, extended thinking, prompt caching, vision.
OpenAI API
GPT setup, structured outputs, vision, and embeddings.
Agentic Workflows
Plan → implement → review loops and CI integration.
RAG for Codebases
Embeddings, doc indexing, pgvector, when to build vs buy.
LLM Observability and Evals
Trace calls, log tokens, run quality evals before shipping.
Tutorials
Quick References
Related
- AI in Web Apps — security, cost, and latency for shipped features
- JavaScript AI-Assisted Development — React/Next.js stack defaults