Official sources¶

Reference · browse · no code

Before this: Resources Primary documentation by vendor. For a ranked, opinionated shortlist with a reason to trust each item, read Resources first.

Every link on this page is a vendor's own documentation. When a detail matters — a parameter, a limit, a price, which model exists this week — go here rather than to any summary, including this site's.

Verified as of 2026-09-02. Every URL was resolved on that date. This page carries no courses-and-certifications section on purpose: the goal is to point you at the source of truth, not at a syllabus.

Anthropic¶

Resource	What it is for
Claude documentation	The API, models, pricing and limits. The starting point for anything Claude
Engineering blog	The highest-density teaching corpus in the field. Context engineering, tool design, harnesses, evals
Claude Code docs	The agentic CLI: skills, subagents, hooks, memory, MCP
Prompt engineering guide	Claude-specific prompting, but most of it generalises
Claude cookbooks	Runnable notebooks. `patterns/agents/` is the workflow taxonomy as code

Microsoft¶

Resource	What it is for
Azure AI Foundry	The umbrella platform. "Azure AI Studio" and most "Azure OpenAI Service" URLs now redirect here
Microsoft Agent Framework	The current .NET and Python agent SDK. GA 1.0 in April 2026; successor to Semantic Kernel and AutoGen
Foundry Agent Service	Hosted agents on Azure, with tools and threads managed for you
Model catalogue	What is actually deployable today, with context limits
Azure AI Search	Hybrid vector and keyword retrieval; the usual RAG layer on Azure
Azure AI Content Safety	Content filtering, prompt shields, groundedness detection
Generative AI for Beginners	A free, self-paced course. Genuinely maintained
AI Agents for Beginners	The agent-focused companion to the above

Semantic Kernel and AutoGen

Both are superseded by Microsoft Agent Framework and are not listed above. Semantic Kernel still receives security fixes into 2027 and existing code keeps working, but new agent work should start on Agent Framework. AutoGen has been in maintenance mode since late 2025. Listicles still recommend both.

OpenAI¶

Resource	What it is for
API documentation	The reference. Note the host: `platform.openai.com` links now redirect here
Models	The live catalogue. Prefer it to any model table, including ours
Agents SDK	The supported agent framework. It replaced Swarm, which was an experiment
Function calling	Tool schemas and the request/response shape the whole industry copied
Embeddings	The `text-embedding-3` models, dimensions and the Matryoshka reduction

Google¶

Resource	What it is for
Gemini API docs	The API, models and limits
Agent Development Kit	Google's open-source agent framework, with its own evaluation tooling
Gemini Enterprise Agent Platform	The managed platform. The old Vertex AI agent URLs land here
AI Studio	Prototyping against Gemini without writing a client
Machine Learning Crash Course	Still the clearest free introduction to learning from data

NVIDIA¶

The layer below the frameworks: what the models actually run on, and the parts you need when you self-host rather than call an API.

Resource	What it is for
NIM	Prebuilt inference microservices. The realistic route to self-hosting a model with an OpenAI-compatible endpoint
NeMo Guardrails	Programmable rails for dialogue, topic and safety control. One of the few guardrail systems with a real specification language
NeMo Guardrails source	Read the rails and the examples; the concepts transfer even if you never deploy it
Technical blog, generative AI	Inference optimisation, quantisation and serving, from the people who build the hardware

Protocols and standards¶

Resource	What it is for
Model Context Protocol	The protocol itself. Check the revision date every time; `2026-07-28` was a breaking redesign
MCP specification	The normative document. The architecture page is better for first contact
MCP Inspector	Debug a server without wiring up a host
A2A protocol	Agent-to-agent delegation. Linux Foundation, not Google, since June 2025
Agent Skills specification	One `SKILL.md` format that runs across several vendors' tooling
OWASP Top 10 for Agentic Applications	The vendor-neutral security taxonomy, each item backed by a real incident

Frameworks and retrieval¶

Resource	What it is for
LangChain	Composable building blocks. Note the host: `python.langchain.com` now redirects
LangGraph	Stateful, graph-based orchestration with cycles and checkpoints
LlamaIndex	Retrieval-first framework with strong ingestion and indexing
CrewAI	Role-based multi-agent orchestration
Microsoft GraphRAG	Graph-based retrieval. Read the indexing cost section before adopting
pgvector	Vector search inside Postgres; the realistic production default
Cohere Embed	Embedding models with explicit input types
MTEB leaderboard	Benchmark comparison for embedding models. A filter, never a verdict

Tooling and community¶

Resource	What it is for
GitHub Copilot docs	Every Copilot surface, including the cloud agent
GitHub MCP server	GitHub's own server. It replaced the archived reference implementation
Hugging Face	Models, datasets and the agents course
Meta Llama	Open-weight models and licences
Ollama	Run models locally. The substrate for every lab on this site