# Capture a URL or preserve one verified media item.

Use `ghostget read` for a non-persistent inspection, pass a URL to Ghostget for durable Markdown capture, or use `ghostget archive` for one authorized, accessible, finite, non-DRM media item with integrity evidence.

This guide describes v0.18.28 and was checked against the public source for that immutable release.

## Choose the result before you run the command

| Command | Result | Persistence |
| --- | --- | --- |
| `ghostget read URL` | Readable page inspection | Does not save the page |
| `ghostget URL` | Durable Markdown page capture and local assets when applicable | Saves to the configured knowledge base |
| `ghostget archive URL` | Verified source media plus requested derivatives, transcript, manifest, and hashes | Saves one archive item |
| `ghostget audio\|video\|transcript URL` | One focused derivative from the same bounded media pipeline | Saves the requested archive view |

## Capture readable web knowledge

```
ghostget read https://example.com
ghostget init ./ghostget-notes
ghostget https://example.com --output ./ghostget-notes/articles
ghostget search "Example Domain" --root ./ghostget-notes --mode exact
```

The saved form is Markdown intended to remain inspectable and searchable in local knowledge. `ghostget context` resolves nearby agent context for a repository path, while `ghostget search` queries the knowledge already kept locally.

Ghostget also exposes the shared `@hraness/kb` URL-intelligence boundary:

```
ghostget url-metadata backfill --root kb
```

The backfill uses a pinned Rust search helper for bounded metadata lookup, writes resumable `url-metadata.json` sidecars beside saved URLs, and performs read-only Archive.today discovery, including archive.is URLs, by default. `--no-archive` disables archive discovery; `--refresh` replaces an existing sidecar after a fresh bounded lookup.

## Preserve media with recomputable integrity

```
ghostget archive https://example.com/authorized-media
ghostget verify path/to/archive-item
```

A completed media item retains the acquired encoded media, privacy-projected provider metadata, requested derivatives and transcripts, a versioned manifest, and SHA-256 integrity records. `ghostget verify` recomputes every recorded artifact hash so the archive can be checked independently after capture.

The unit of work is exactly one accessible, finite item. Ghostget rejects playlists, live streams, affirmative DRM, and unsupported authentication rather than silently weakening the boundary. It does not bypass authentication, payment, access controls, or DRM.

## Local browser work is bounded

Fresh and profile-backed page capture share at most two locally owned browser acquisitions across Ghostget processes using the same state home. Admission consumes the capture timeout and remains held through browser, proxy, process, and isolation cleanup.

An explicit `--cdp` or `--browser-live` attachment does not launch a Ghostget-owned browser and does not consume one of those two slots. Ambiguous same-boot admission state stays occupied rather than allowing capacity to exceed the bound. Run `ghostget doctor --json` for the exact state home and recovery evidence.

## Email thread clipping is a separate authenticated path

A Gmail `messaging.list` result includes a provider-derived `threadUrl`. Pass that URL to `ghostget read` or `ghostget clip` with the same Gmail auth locator. Gmail clips default to private Ghostget state; `--output <directory>` is the explicit plaintext export boundary.

The default and `--media all` include every MIME attachment. `--media none` omits attachment bytes, while `--media images` is rejected because it would misrepresent non-image files. Attachment objects are content-addressed, integrity-recorded, and stored under deterministic SHA-256 names. See the [provider support directory](https://ghostget.com/docs/reference/provider-capabilities/) for available inbox, search, thread, and contact actions.

## Keep the capture boundary explicit

Install and verify the CLI with the [getting started guide](https://ghostget.com/docs/tutorials/getting-started/). For account-bound reads, review [local custody and provider contracts](https://ghostget.com/docs/explanation/security-model/) before adding credentials.