Practical Bun Docs: Toolchain, Cloud, and AI Development

Bun is a JavaScript and TypeScript toolchain. One executable provides a runtime, package manager, test runner, and bundler. This field guide complements the official Bun API reference with shorter learning paths, explicit decision boundaries, and context that agents can consume directly.

Pick a reading path

I am new to Bun

Install it, build the mental model, and finish a small project.

I am migrating a Node.js project

Adopt one boundary at a time and verify compatibility.

I am an AI agent

Route the task first, then load only the pages you need.

I am deploying to cloud

Distinguish the real runtime across containers, serverless, and edge.

I am building an AI app

Streaming, tool safety, RAG, observability, and evaluations.

I need data and storage

Bun.SQL, Redis, S3, Drizzle, and production connection boundaries.

I need a production baseline

TypeScript, CI, supply chain, environment, and release gates.

I need a command

A compact map of commands, config files, and diagnostics.

Capability and task map

Goal	First command	Guide
Execute `.js`, `.ts`, or `.tsx`	`bun run <file>`	Runtime
Install npm packages	`bun install` / `bun add`	Package manager
Run Jest-style tests	`bun test`	Test runner
Produce browser or server artifacts	`bun build`	Bundler
Organize applications and shared packages	`bun --filter`	Monorepos and workspaces
Use SQL, Redis, or S3	`Bun.SQL` / `RedisClient` / `S3Client`	Data and storage
Establish reproducible release gates	`bun ci`	Production engineering baseline
Choose frameworks, ORMs, and tooling	List constraints first	Ecosystem decisions

Version language

These guides use 1.x for the stable major line instead of baking a fast-expiring patch number into examples. Run bun --version and consult the matching official release notes when behavior depends on a precise version.

Two audiences, one source of truth

Newcomer pages explain intent and mental models before commands, expected results, and common failure modes.
Agent pages keep stable headings, explicit prerequisites, decision tables, and clear “do not guess” boundaries.
Append .md to any docs URL for Markdown. The site also publishes /en/llms.txt and /en/llms-full.txt.