blueprintslisted

# Blueprints A blueprint records how this project builds a particular kind of thing: the shape, the contracts, the failure modes, and why each choice was made. Following one keeps a new feature consistent with what is already here — which usually matters more than whether any single decision inside it was the best available. ## Choosing one Match the task against the *when it applies* column below, then read that file in full. If nothing matches, there is no blueprint for this work. Say so and carry on normally: a blueprint stretched over a task it was not written for imports decisions that were made for a different problem, which is worse than having no blueprint at all. | Blueprint | When it applies | |---|---| | [Agent-to-agent (A2A)](blueprints/a2a-agent-to-agent.md) | Handing an outcome to an autonomous agent you do not own — another team's or another company's — by specifying the goal rather than the call | | [AG-UI frontend streaming](blueprints/ag-ui-frontend-streaming.md) | The interface has to show what the agent is doing while it works: tokens, tool calls, thinking traces, interruptions | | [Agent loop](blueprints/agent-loop.md) | Building the loop that calls a model, runs the tools it asks for and repeats until it stops; or an agent runs away, loops, or will not stop on its own | | [Agent protocols overview](blueprints/agent-protocols-overview.md) | Choosing which wire protocol fits — tools, agent-to-agent delegation, frontend streaming — when an agent has to