# scrapewright

> Give it a URL, get structured rows. A model writes the parser once; replays are free.

- name: io.github.Ozymandias-Owens-2/scrapewright
- version: 1.0.3
- connection class: R1 (autonomous)
- trust: 97/100
- quality: 73/100 (needs work)
- owner: not claimed
- descriptor: https://api.protogrid.dev/v1/servers/io.github.Ozymandias-Owens-2%2Fscrapewright
- tools: https://api.protogrid.dev/v1/servers/io.github.Ozymandias-Owens-2%2Fscrapewright/tools

**An agent can connect on its own if it already holds `X_API_KEY`, `X_API_KEY`.** Remote endpoint behind a static API key or header. The registry never sees the value; the agent fills the placeholder.

## Connection (mcpServers)

```json
{
  "mcpServers": {
    "scrapewright": {
      "type": "http",
      "url": "https://scrapewright.app/mcp",
      "headers": {
        "X-API-Key": "${X_API_KEY}",
        "Authorization": "${X_API_KEY}"
      }
    }
  }
}
```

Placeholders to fill: `${X_API_KEY}`, `${X_API_KEY}`.

## Trust

- hygiene: 100
- liveness: 100
- freshness: 100
- provenance: 90
- drivers: +repository +namespace-matches-repo +website +reachable +uptime-99% +updated-21d-ago
- Derived from observable signals (official registry feed and our own credential-free probes); no code audit performed.

## Quality 73/100 (needs work)

### protocol: 88 (weight 25)

- pass `protocol.modern`: supports 2026-07-28 (server/discover)
- pass `protocol.stateless`: answers without a session
- pass `protocol.transport`: streamable HTTP
- warn `protocol.list_ttl`: tools/list declares ttlMs 0, so clients re-fetch it on every use

### authorization: n/a (weight 15)

- n/a `auth.prm`, `auth.as_metadata`, `auth.cimd`: no remote uses OAuth
- n/a `auth.secret_in_url`: no templated URL

### tool hygiene: 71 (weight 30)

- pass `tools.descriptions`: every tool has a description
- pass `tools.description_length`: descriptions are concise
- pass `tools.schemas`: every tool has an object input schema
- fail `tools.annotations`: no tool declares readOnlyHint or destructiveHint, so clients cannot tell safe calls from risky ones
- fail `tools.directory_hints`: no tool declares all four of readOnlyHint, destructiveHint, idempotentHint and openWorldHint; missing: extract_page (readOnlyHint, destructiveHint, idempotentHint, openWorldHint); crawl_site (readOnlyHint, destructiveHint, idempotentHint, openWorldHint); crawl_status (readOnlyHint, destructiveHint, idempotentHint, openWorldHint); …
- pass `tools.token_cost`: about 659 tokens to load every tool
- pass `tools.api_dump`: tools are not a one-to-one API dump

### stability: 50 (weight 15)

- pass `stability.changes`: 2 tool changes in 30 days
- fail `stability.rug_pull`: 1 tools kept their name but changed most of their description in 30 days; review before trusting them

### dependencies: n/a (weight 15)

- n/a `deps.known_vulns`, `deps.mcp_sdk_version`, `deps.resolvable`: deps.dev does not know scrapewright@1.0.3

- tools: 5, about 659 tokens to load them all
- tool set hash: acb7fad529785ea3e179215d54be4e02
- tools last changed: 2026-10-01T18:12:26.753Z
- badge: [![protogrid quality](https://protogrid.dev/badge/io.github.Ozymandias-Owens-2/scrapewright.svg)](https://protogrid.dev/servers/io.github.Ozymandias-Owens-2/scrapewright)
- Checks run on what our credential-free, read-only probes observe; tools are never called and no code audit is performed.

### Tool changes

- 2026-10-01 `extract_page` changed (description, inputSchema)
  - before: Extract structured data from ONE page. ``fields`` declares your own schema, e.g. ["title", "salary:number", "tags:list"]; omit it for the product schema. First call on a new site compiles a recipe (300 credits); later calls replay it for 1 credit per row.
  - after: Extract structured data from ONE page. ``fields`` declares your own schema, e.g. ["title", "salary:number", "tags:list"]; omit it for the product schema. First call on a new site compiles a recipe (300 credits); later calls replay it for 1 credit per row. A site nobody could read is remembered for a week and answered without calling the model, so re-reading a dead page costs nothing. `retry` asks anyway.
- 2026-10-01 `crawl_site` changed (description, inputSchema) **meaning changed**
  - before: Walk a site from one listing URL and extract every item. Waits up to four minutes; a longer crawl returns a job_id to pass to crawl_status.
  - after: Walk a site and extract every item. Waits up to four minutes; a longer crawl returns a job_id to pass to crawl_status. `mode` picks how the items are found. "rows" (the default) reads every card on the listing itself and follows its pagination -- right for search results, categories and tables. "links" follows each card into its own page, for sites where an item has one. "like" starts from one example item page, given as `like_page`, and finds every other page shaped like it; pass `listing_url` as well when you know the stock or results page, because inferring it is the part that fails.

Full history: https://api.protogrid.dev/v1/servers/io.github.Ozymandias-Owens-2%2Fscrapewright/changes

## Dependencies

- pypi scrapewright@1.0.3: unknown

Dependency graphs from deps.dev; advisories from OSV.dev, including the GitHub Advisory Database and the PyPI Advisory Database; all CC BY 4.0.

## Remotes

- https://scrapewright.app/mcp (streamable-http, auth api_key ${X_API_KEY}, reachable true, uptime30d 0.99159664)

## Packages

- pypi scrapewright@1.0.3 secrets: ANTHROPIC_API_KEY

## Tools (5)

- `account`: Credits left and this month's usage for the key in use.
- `crawl_site`: Walk a site and extract every item. Waits up to four minutes; a
        longer crawl returns a job_id to pass to crawl_status.

        `mode` picks how the items are found. "rows" (the default) reads every
        card on the listing itself and follows its pagination -- right for
        search results, categories and tables. "links" follows each card into
        its own page, for sites where an item has one. "like" starts from one
        example item page, given as `like_page`, and finds every other page
        shaped like it; pass `listing_url` as well when you know the stock or
        results page, because inferring it is the part that fails.
        
- `crawl_status`: Fetch a crawl that outlived its call.
- `detect_site`: Report what platform a site runs on and which strategy to use.
        Cheap; call it before a large job.
- `extract_page`: Extract structured data from ONE page. ``fields`` declares your own
        schema, e.g. ["title", "salary:number", "tags:list"]; omit it for the
        product schema. First call on a new site compiles a recipe (300
        credits); later calls replay it for 1 credit per row.

        A site nobody could read is remembered for a week and answered
        without calling the model, so re-reading a dead page costs nothing.
        `retry` asks anyway.