bmad-technical-research

Tell the agent you want to produce a technical research report and state your topic, such as a framework comparison or architecture question. The workflow then guides you through scope confirmation, analysis steps, and a final synthesis, asking for confirmation at each stage.

It covers any technology or architecture topic, including programming languages, frameworks, databases, cloud platforms, API design, integration patterns, and implementation practices. The scope is confirmed with you in the first step before research begins.

Yes, web search is a mandatory prerequisite. The workflow aborts if web search is unavailable, because every technical claim must be verified against current live sources with URL citations rather than relying on training data.

The report is saved as a Markdown file under the planning artifacts directory in a research subfolder, named with the topic slug and date. It starts from a template and is appended to progressively through each workflow step.

Yes, a customize.toml file supports activation step overrides, persistent facts loaded from project files, and an on-complete hook. Team and user-level TOML overrides merge with the defaults using structural merge rules.

It depends entirely on web search availability and source quality, so niche topics with little public documentation may yield lower-confidence findings. It also requires interactive confirmation at each step, so it cannot run fully unattended.