Say exactly what can leave the device, because "only" was wrong

The README claimed the paraphrase check was the only feature that sends text
anywhere. There are four: paraphrase, the perplexity measurement, the live rewrite
with your own key — the key stays put, the text does not — and the optional web
spot-check for a distinctive phrase.

Every one is off until switched on and disclosed at the moment of choosing, so the
spirit of the claim held. The letter did not, and this is the one sentence in the
project a reader is entitled to take literally. A privacy promise wider than the code
is the sort of thing this project exists to complain about elsewhere.

The MCP README listed nine tools and said "the first seven"; write_report shipped in
0.4.0 and eight of the ten now run offline.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

Author: peopleworks

README.md

@@ -12,7 +12,7 @@
 [![NuGet MCP](https://img.shields.io/nuget/v/SignsOfAI.Mcp?logo=nuget&label=MCP%20server)](https://www.nuget.org/packages/SignsOfAI.Mcp)
 [![Available on CodeGuilds](https://img.shields.io/badge/Available_on-CodeGuilds-6366f1?logo=data:image/svg%2bxml;base64,PHN2ZyB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciIHZpZXdCb3g9IjAgMCAyNCAyNCI+PHBhdGggZmlsbD0id2hpdGUiIGQ9Ik0xMiAyTDIgN2wxMCA1IDEwLTV6TTIgMTdsMTAgNSAxMC01TTIgMTJsMTAgNSAxMC01Ii8+PC9zdmc+)](https://codeguilds.dev/packages/signsofai)
 
-**[Try the live demo →](https://peopleworks.github.io/SignsofAI/)** — English & Spanish, runs 100% in your browser. No signup, and nothing leaves your device.
+**[Try the live demo →](https://peopleworks.github.io/SignsofAI/)** — English & Spanish, runs in your browser. No signup, and the analysis uploads nothing.
 
 **[Download the Windows app →](https://github.com/peopleworks/SignsofAI/releases?q=desktop&expanded=true)** — the same tool in a window. Nothing to install alongside it: the .NET runtime is bundled.
 
@@ -31,8 +31,19 @@ A free, privacy-first toolkit for **academic and writing integrity**. It does tw
    and surfaces the passages they share — verbatim copies, **reworded paraphrases** (even across
    languages), and a whole-cohort overview — as **evidence a human judges**. Not a black-box verdict.
 
-> 🔒 **Almost everything runs 100% in your browser. Your documents never leave your device.**
-> The only exception is the optional paraphrase check, which is strictly opt-in and clearly disclosed.
+> 🔒 **The analysis runs entirely in your browser, and nothing is uploaded to run it.** No account, no
+> telemetry, no server that sees your text.
+>
+> Four features can send text off the device, and **not one of them runs unless you turn it on**, each
+> disclosed in the interface at the moment you choose it: the **paraphrase check** and the
+> **perplexity measurement** (both call a server you or we host), the **live rewrite** when you supply
+> your own API key — the key stays on your device, the text goes to the provider you picked — and the
+> optional **web spot-check** for a distinctive phrase, which exists only if the operator configured a
+> search provider.
+>
+> Everything else — every rule, the score, the character scan, the citation cross-check, the writer
+> baseline, the report — is computed locally and stays there. In the desktop app, the perplexity
+> measurement is local too.
 
 Built with **.NET 10** and **Blazor WebAssembly** by **Pedro Hernández (PeopleWorks)**, [Microsoft MVP for .NET](https://mvp.microsoft.com/en-US/mvp/profile/24060a02-dbc6-44ec-bca5-c213ff9835c5) — for the .NET and Microsoft developer community, *por y para la comunidad educativa*.
 

src/SignsOfAI.Mcp/README.md

@@ -18,11 +18,13 @@ Built on the official [`ModelContextProtocol`](https://www.nuget.org/packages/Mo
 | `inspect_characters` | Characters typing cannot produce — zero-width, homoglyphs, hidden tag characters — with the exact line and column of each | 🖥️ offline |
 | `check_citations` | Where a document contradicts its **own** reference list: a source cited but never listed, one DOI on two works, a year that has not happened | 🖥️ offline |
 | `compare_to_baseline` | How a piece sits against the same person's earlier work (Burrows's Delta) — measured on **their** variation, never a verdict about who wrote it | 🖥️ offline |
+| `write_report` | The whole analysis as a document a person can keep, forward, or take to a committee — with the tool's own error rate printed on it | 🖥️ offline |
 | `measure_predictability` | Perplexity — how predictable/generic a model finds the phrasing | ☁️ server |
 | `check_paraphrase` | Reworded/translated copies via sentence embeddings (EmbeddingGemma) | ☁️ server |
 
-The first seven run **entirely on the machine** — the text never leaves it. The last two **send the text**
-to the SignsOfAI server (their descriptions disclose this); see [Server tools](#server-tools-optional).
+Eight of the ten run **entirely on the machine** — the text never leaves it. The two marked ☁️ **send
+the text** to the SignsOfAI server (their descriptions disclose this); see
+[Server tools](#server-tools-optional).
 
 ## Install