Memo

Legal memos are written for one reader and forwarded to many. The forwards never land. A client gets a memo full of case law they don't understand. A junior gets a memo that doesn't tell them what to do next. A senior gets a memo that buries the risk three pages in.

Memo solves this by writing the right shape for each reader. Three audiences, one output per audience.

Calls the Anthropic API with your own key. No backend, no database, no logs. Your memo and your key stay in your browser — they never reach a Memo server, because there isn't one.

The client gets a plain-English email with the decision at the top. The junior gets a research and drafting brief — what to read, what to draft, what to chase. The senior gets a risk and sign-off note - top risks, what to escalate, what's settled.

Every factual claim in the output cites the source paragraph it came from, with a confidence label. High when the memo states it directly. Med when synthesised across paragraphs. Low when it's professional judgment. If the tool can't cite a claim, it won't make it.

The app is a demo. The product is a Claude skill - a portable Markdown file with a prompt and a contract. Anyone with Claude can run the skill without this app. We think the unit of legal-tech distribution is a skill, not an app.

collaborators for legal ai ideas, dms open!