Add Scrapiq remote MCP server (URL to clean text/markdown/JSON for RAG)

[NG-PR0JECT]

What this adds

Scrapiq — a remote MCP server that turns any URL into clean text, markdown, or JSON for LLM and RAG pipelines.

• Type: remote, streamable-http, anonymous (no OAuth, no API key, no account)
• Endpoint: https://scrapiq.io/mcp
• Source: https://github.com/NG-PR0JECT/scrapiq-mcp-server (MIT)
• Docs: https://scrapiq.io/docs

Why it is useful in the Toolkit

Feeding raw HTML into retrieval silently degrades retrieval quality and inflates token spend; most scrapers either drag in navigation/ads/boilerplate or fail silently and return an empty string with no error. Scrapiq exposes one tool, scrapiq_extract(url, format, max_chars), which returns title, content, links and metadata for a page — no LLM in the loop, and no extraction keys to configure.

It pairs with the local filesystem/git servers: an agent can pull a page (release notes, a docs page, a competitor's pricing page) into clean markdown and then write it to disk, with no API credentials to paste into the Toolkit UI.

Verification performed

• initialize → protocol 2025-06-18; tools/list → 1 tool (scrapiq_extract) against the public endpoint, verified with the official MCP Python SDK over plain HTTPS.
• POST /mcp returns 200/202 (notifications), GET /mcp returns 405 — stateless by design, which is expected for a remote server.
• Public benchmark vs trafilatura / readability-lxml / MarkItDown on 23 live pages (raw output, method and caveats published): https://github.com/NG-PR0JECT/scrapiq-bench

Notes

• tools.json is [] because tools are discovered dynamically from the remote endpoint.
• No secrets or environment variables are needed to enable the server.

[NG-PR0JECT]

Gentle ping — the two runs on this branch (.github/workflows/ci.yaml and Security Review Trigger) are both conclusion: action_required with 0 s runtime, which I read as the first-time-contributor approval gate rather than a problem with the submission. Nothing else here is waiting on me:

• the PR adds three files under servers/scrapiq/ (readme.md, server.yaml, tools.json), and is MERGEABLE;
• npx prettier --check passes on all three exactly as committed;
• source.commit fb3e050 is reachable from NG-PR0JECT/scrapiq-mcp-server@main, and that repository is MIT-licensed;
• the remote declared in server.yaml answers right now: POST https://scrapiq.io/mcp with an initialize returns protocol 2025-03-26 and serverInfo scrapiq-mcp-server 0.2.0, and tools/list returns scrapiq_extract;
• dynamic: {tools: true} is set, so CI never has to reach the endpoint, and tools.json is [] accordingly.

Anything needed from my side for the runs to be approved, or a maintainer action to wait on? Happy to change the category or tag set if the catalogue prefers something else.