Memo
Tell me what you need. I'll write the prompt.
Paste what you have, de-identified: the fact pattern, the notice, the client's question in their words. Ledger writes the finished, grounded prompt and runs it. Then it shows you every authority it would have had to open to be sure, and marks what it could not check.
What you paste goes to Anthropic's Claude API to write the draft, and nowhere else. It is not used to train a model, and it is not stored on our servers unless you sign in and sync. Where your data goes →
Every output is a draft for the preparer who signs to check. It states no figure, threshold or rate it was not given, marks every authority VERIFY until you have opened it, and quotes a deadline as the notice prints it rather than computing one. Not tax, legal or accounting advice to any client. Nothing identifiable belongs in the paste: removing names reduces risk, but it may not de-identify an unusual fact pattern, and whether a de-identified question falls outside section 7216 is unresolved. This surface is not a safe harbor.