diff --git a/README.md b/README.md index a2b46a6..f9aa8c9 100644 --- a/README.md +++ b/README.md @@ -14,14 +14,17 @@ One file. Zero runtime dependencies. No network requests. ## Package -The npm package name is `@yrstm/mantis`. +The package name is `@yrstm/mantis`. It is not published to the npm registry yet, so install the +current public build from its pinned GitHub commit: ```sh -npm install @yrstm/mantis +npm install https://github.com/yrstm/mantis/archive/9e462696990449979b309886d3581dd3201c2b17.tar.gz ``` The official Mantis browser extension is packaged separately from this open-source library repo. -The store link will be added here when it is ready. +See the [Mantis Extension page](https://yrstm.github.io/mantis/extension/) for the capture flow, +privacy boundary, current limitations, and release status. Store links will be added there when +they are ready. ```js const article = Mantis.extract(document); diff --git a/docs/extension/index.html b/docs/extension/index.html new file mode 100644 index 0000000..92d18c7 --- /dev/null +++ b/docs/extension/index.html @@ -0,0 +1,952 @@ + + + + + + + + + + + + + Mantis Extension — The web, in clean Markdown + + + +
+
+ + + Mantis + Extension + + +
+
+ +
+
+
+
+

Chrome + Safari · pre-release

+

The web, in clean Markdown.

+

Mantis Extension turns the rendered page—or just your selection—into structured Markdown you can paste straight into an agent.

+
+ See release status + See how it works +
+
    +
  • No account
    • +
  • User-triggered
    • +
  • No external fallback
    • +
+
+ + +
+
+ +
+
+
+

One-click capture

+

From busy page to useful context.

+

Mantis works on the DOM already rendered in your tab. It keeps the article-shaped content and leaves navigation, banners and repeated chrome behind.

+
+
+
+ 1 +

Open the page

+

Use the content already visible in your browser, including client-rendered pages and signed-in state.

+
+
+ 2 +

Click Mantis

+

Capture the active selection when there is one; otherwise extract the page's main readable content.

+
+
+ 3 +

Paste the Markdown

+

Preview the result, see confidence and warnings, then paste structured context into the tool you choose.

+
+
+
+
+ +
+
+
+

Made for agent context

+

Structure without the HTML weight.

+

The extension uses the open-source Mantis extraction engine and packages it into a focused browser workflow.

+
+
+
+ +

Rendered-page capture

+

Works after JavaScript rendering instead of fetching a second, potentially different copy.

+
+
+ +

Selection aware

+

A useful selection takes priority, making it easy to capture only the passage you need.

+
+
+ +

Markdown structure

+

Preserves headings, lists, quotes, code, links, content images and data tables in reading order.

+
+
+ +

Context signals

+

Adds source metadata, capture mode, confidence, warnings, hashes and a rough token estimate.

+
+
+ +

In-page preview

+

Shows a small isolated panel on the page, with a manual-copy fallback when clipboard access is blocked.

+
+
+ +

Agent safety marker

+

Marks captured web content as untrusted data so downstream tools can keep it separate from instructions.

+
+
+
+
+ +
+
+
+

Privacy by boundary

+

The core capture stays in your browser.

+

The normal one-click path needs no Mantis account, analytics service or Mantis backend.

+
    +
  • Runs only when you click the browser action.
    • +
  • Reads the active HTTP(S) tab or selection for that capture.
    • +
  • Converts locally and writes the result to your clipboard.
    • +
  • Ships its extraction code inside the extension package.
    • +
+
+ +
+
+ +
+
+
+

Availability

+

Built for Chrome and Safari. Not in stores yet.

+

The browser packages exist and build locally. Store submission, signing and final clean-profile release QA are intentionally deferred.

+
+
+
+
+
Chrome & Chromium
+ Pre-release +
+

The Manifest V3 package builds and can be loaded unpacked for internal testing.

+
    +
  • Chrome Web Store submission: deferred
    • +
  • Public install URL: not available yet
    • +
  • Local-only privacy boundary: verified
    • +
+
+
+
+
Safari for macOS
+ Pre-release +
+

The Safari Web Extension wrapper is generated from the same extension resources.

+
    +
  • Signed App Store submission: deferred
    • +
  • Public install URL: not available yet
    • +
  • Unsigned local wrapper build: verified
    • +
+
+
+
+ +
This is the canonical extension page. Store distribution is planned for a later release; real Chrome and Safari install buttons will replace this status after their listings are approved.
+
+
+
+ +
+
+
+

Know the edges

+

What the first release will—and will not—do.

+

Mantis is a heuristic content extractor, not a verbatim archive or a browser automation service.

+
+
+
+

Does it capture every page?

+

No. Browser-internal pages, extension stores and other restricted URLs cannot be injected. Unusual layouts can produce partial or low-confidence output.

+
+
+

Does it include hidden or unloaded content?

+

It captures the DOM currently rendered in the tab. Content that has not loaded, lives in inaccessible frames or is hidden is outside that snapshot.

+
+
+

Are screenshots and diagrams understood?

+

The browser extension keeps image references and alt text; it does not run vision over diagrams. Image/OCR capture is a separate Mantis workflow.

+
+
+

Is captured text automatically safe for an agent?

+

No web content is inherently trusted. Mantis adds a safety marker, but the receiving agent must still treat the Markdown as data rather than instructions.

+
+
+
+
+ +
+
+
+

Want the engine today?

+

Explore the open-source Mantis library, its test corpus and the live paste converter.

+
+ Open the Mantis library page +
+
+
+ + + + diff --git a/docs/index.html b/docs/index.html index 24c94dd..8097f35 100644 --- a/docs/index.html +++ b/docs/index.html @@ -3,6 +3,8 @@ + + Mantis @@ -248,6 +270,7 @@
MMantis