Most agent projects make a few big choices early: which framework to build on, which model to run, and how to give the agent its tools. Each choice is hard to undo later, and marketing pages rarely say where a tool is weak.
These guides compare the options with working code and plain trade-offs. The framework comparison sets the main Python agent frameworks side by side, including LangGraph, Pydantic AI, the OpenAI Agents SDK, CrewAI and Promptise Foundry, and says which fits which kind of project. The model guide explains what matters for tool calling and how to test candidates on your own workload. The MCP versus CLI guide covers when a protocol server is worth it and when a command-line tool is simpler and cheaper.
The code in these guides runs on Promptise Foundry, an open-source Python framework for AI agents and MCP servers.