@preserve-cli/seed

Seed brand profiles, images, products, proof, glossary, and overview documents into a Preserve workspace from any public marketing website.

Prerequisites

• An existing Preserve workspace
• A workspace API key with seed/publish scope (brand.seed), created by a workspace owner or admin

Usage

npx @preserve-cli/seed \
  --url https://example.com \
  --slug my-workspace \
  --auth-key pk_live_…

Verify published workspace knowledge:

npx @preserve-cli/seed verify \
  --slug my-workspace \
  --auth-key pk_live_…

Dry-run scrape only (writes local cache, no publish):

npx @preserve-cli/seed \
  --url https://example.com \
  --slug my-workspace \
  --auth-key pk_live_… \
  --dry-run

Environment variables

• PRESERVE_API_KEY — fallback for --auth-key
• PRESERVE_BASE_URL — fallback for --base-url (default: https://mcp.preserve.design)

Commands

| Command | Description | |---------|-------------| | preserve-seed (default) | Crawl a website and publish full workspace knowledge | | preserve-seed verify | Read back published brand, assets, products, glossary, and proof via MCP |

Flags

| Flag | Description | |------|-------------| | --url | Website to scrape (required unless --skip-scrape) | | --slug | Workspace slug | | --auth-key | pk_live_ key with brand.seed scope | | --base-url | MCP host (default https://mcp.preserve.design) | | --max-pages | Maximum marketing pages to crawl (default 40) | | --include-blog | Include /blog pages in the crawl | | --dry-run | Scrape only; write local cache, no publish | | --skip-scrape | Publish from existing local cache | | --output-dir | Cache directory (default ./.preserve/{slug}/) |

What gets scraped

The CLI discovers pages via sitemap.xml, homepage nav/footer links, and priority marketing paths (/about*, /platform*, /product*, /pricing, /glossary, /case-study*, etc.).

It publishes:

• Brand — company name, voice, colors, typography (from meta tags, JSON-LD, and CSS)
• Assets — favicon, OG image, primary logo (customer/partner logos filtered out)
• Products — one product per matched marketing/product page
• Proof — case studies, testimonials, homepage metrics, FAQ claims
• Glossary — terms from /glossary plus company name
• Overview — markdown company overview document

Local cache

After scraping, files are written to .preserve/{slug}/:

pages.json
brand.json
assets-manifest.json
products.json
proof.json
glossary.json
overview.md

Use --skip-scrape to re-publish from cache without re-crawling.

Local development

From the monorepo root:

pnpm --filter @preserve-cli/seed build
pnpm preserve -- --url https://www.docket.io --slug docket-inc --auth-key pk_live_… --dry-run
pnpm preserve verify -- --slug docket-inc --auth-key pk_live_…

Publish to npm

The package is standalone (no monorepo workspace dependencies). From the repo root:

pnpm --filter @preserve-cli/seed pack:check   # build + validate tarball
pnpm preserve:publish                         # npm publish --access public

Requires npm login with publish access to @preserve-cli.