Architecture Decision Scribe

For software and service support teams investigating customer-reported errors. Give this agent a product domain and an error message; it researches public documentation, release notes, and status pages with Google Grounding, then delivers cited related findings, clearly labeled possible causes, and targeted questions to enrich a support ticket without claiming access to private logs or customer accounts.

Tell it what you want to create (a blog post, landing page or product page) plus the country and language, and it picks the keyword worth targeting with the evidence behind it. It discovers candidates with Google keyword data, verifies volume, difficulty, CPC and intent, checks who actually ranks on Google for the primary term, then returns a structured report: primary and supporting keywords, intent clusters, SERP evidence with content gaps, and an SEO-ready title, meta description, H1 and outline. Every number comes from a tool call in the run. Nothing is estimated.

For online store owners and marketplace sellers who want to improve abandoned-cart follow-up without eroding contribution margin. Paste aggregate cart counts, shipping charges and costs, product costs or gross margins, and recovery results by segment. The agent checks the inputs, uses code to compare recovery and shipping economics, models discount and shipping-offer scenarios, and returns a ranked table of financially bounded follow-up offers to test with measurement and guardrails.