Build

Skills, engineering guides and model APIs for practical AI work.

For builders: Claude Code and agent operations, prompt and context engineering, model APIs, frameworks, evals, and shipping to production.

Agent Engineering

Claude Agent SDKClaude Code as a library — build your own agents on the same harness and tool set

Building Effective AgentsAnthropic's canonical guide: workflows vs agents, and when you don't need an agent at all

Agent Skills (Engineering Blog)Why skills beat fine-tuning for repeatable work — the design behind the skills system

Prompt & Context Engineering

Anthropic Prompt EngineeringOfficial techniques: clear instructions, examples, XML tags, chain-of-thought, prefills

Effective Context EngineeringBeyond prompts: managing what an agent sees across long tasks — compaction, memory, tool results

OpenAI Prompt EngineeringOpenAI's official prompting guide for the GPT family

Prompting GuideModel-agnostic prompt engineering techniques, with research citations

Model APIs & Protocols

Claude API DocsMessages API, tool use, structured outputs, prompt caching, batches — see the APIs tab for endpoints

OpenAI API DocsGPT models, images, speech, and the chat/completions format others emulate

Google Gemini APIGemini models via AI Studio, with a generous free tier for prototyping

Model Context ProtocolThe open standard for connecting agents to tools and data — spec, quickstarts, SDKs

Frameworks & SDKs

LlamaIndexRAG pipelines and data connectors for LLM applications

Vercel AI SDKStreaming AI apps with React/Next.js — one interface across providers

Hugging Face TransformersLoad, run, and fine-tune open-weight models in Python

Evals & Reliability

promptfooOpen-source eval harness: test prompts and agents against assertions in CI

LangSmithTracing, evals, and monitoring for LLM applications

BraintrustEval-driven development: score outputs, compare models, catch regressions

Run & Ship

OllamaRun open-weight LLMs locally with one command

vLLM DocsProduction LLM serving — the standard for self-hosting open models at scale

GradioTurn a model or agent into a shareable web demo in minutes

Hugging Face HubModels, datasets, and Spaces — where open AI actually ships

ReplicateRun and deploy ML models behind a simple API

Agent internals & research

Frameworks and research implementations for studying agent architecture, memory, and evaluation.

smolagentsHugging Face

Minimal agent library (~1,000 core lines) where agents write Python code to act instead of emitting JSON tool calls. The cleanest codebase to learn agent internals from.

LettaLetta (MemGPT)

Agent server built around the MemGPT research: agents with self-editing long-term memory that persist across sessions, exposed through a clean API.

SWE-agentPrinceton / SWE-agent team

The academic system behind SWE-bench: turns a language model into a software engineering agent that fixes real GitHub issues, with a carefully designed agent-computer interface.