The whole-topic guide hubs — start here to answer a build decision end to end. Each hub is an editor-ordered map of the comparisons, explainers and buyer's guides for one part of the AI-agent stack, read in order.
What MCP is vs function calling, its primitives, building and securing servers, the 2026 stateless spec, and who controls the protocol.
Open the guide →LangGraph, CrewAI, the OpenAI and Claude agent SDKs, Pydantic AI and the rest — how the orchestration models actually differ.
Open the guide →Vector databases, embeddings, chunking, rerankers, hybrid and graph RAG — the retrieval stack behind a grounded agent.
Open the guide →Short- vs long-term memory, the memory frameworks (Mem0, Zep, Letta), and how agents remember across sessions.
Open the guide →Serving engines, quantization, KV-cache and the token economics that decide what an agent costs to run at scale.
Open the guide →Eval frameworks, LLM-as-judge, observability and tracing — how to know whether an agent actually works.
Open the guide →Prompt injection, tool poisoning, the confused-deputy trap, sandboxing and the OWASP agent and MCP threat surface.
Open the guide →The IDE assistants and CLI agents, edit formats and fast-apply, AGENTS.md, AI code review and the coding-agent harness.
Open the guide →Claude vs GPT vs Gemini, the frontier tiers, the open-weight field, small models, and the token economics that move the bill.
Open the guide →