dnb-mcp 0.1.0__tar.gz

dnb_mcp-0.1.0/.github/workflows/publish.yml

@@ -0,0 +1,20 @@
+name: Publish to PyPI
+
+on:
+  push:
+    tags: ["v*"]
+
+jobs:
+  publish:
+    runs-on: ubuntu-latest
+    permissions:
+      id-token: write # required for PyPI trusted publishing (OIDC)
+      contents: read
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+      - run: pip install build
+      - run: python -m build
+      - uses: pypa/gh-action-pypi-publish@release/v1

dnb_mcp-0.1.0/.github/workflows/registry.yml

@@ -0,0 +1,61 @@
+name: Publish to MCP Registry
+
+# Official MCP Registry (registry.modelcontextprotocol.io) via GitHub OIDC.
+# No secret, no interactive login: the runner's OIDC token proves ownership
+# of the io.github.mohammad-abbasi-sh/* namespace.
+# Refs: https://github.com/modelcontextprotocol/registry
+
+on:
+  push:
+    tags: ["v*"]
+  workflow_dispatch:
+
+jobs:
+  publish:
+    runs-on: ubuntu-latest
+    permissions:
+      id-token: write # required for OIDC -> MCP Registry
+      contents: read
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Sync server.json version to tag
+        # Only sync when running on a version tag (v*). On workflow_dispatch
+        # (ref = refs/heads/main) keep the version committed in server.json.
+        run: |
+          VER=""
+          if [[ "$GITHUB_REF" == refs/tags/* ]]; then
+            TAG="${GITHUB_REF#refs/tags/}"
+            VER="${TAG#v}"
+          fi
+          if [ -n "$VER" ]; then
+            python3 - "$VER" << 'EOF'
+          import json, sys
+          ver = sys.argv[1]
+          d = json.load(open('server.json'))
+          d['version'] = ver
+          for pkg in d.get('packages', []):
+              pkg['version'] = ver
+          json.dump(d, open('server.json', 'w'), indent=2, ensure_ascii=False)
+          print('server.json version ->', ver)
+          EOF
+          else
+            echo "No version tag ($GITHUB_REF); keeping server.json version as-is"
+          fi
+
+      - name: Install mcp-publisher
+        run: |
+          curl -sL "https://github.com/modelcontextprotocol/registry/releases/download/v1.8.7/mcp-publisher_linux_amd64.tar.gz" -o /tmp/mcp-publisher.tar.gz
+          tar -xzf /tmp/mcp-publisher.tar.gz -C /tmp
+          sudo mv /tmp/mcp-publisher /usr/local/bin/mcp-publisher
+          mcp-publisher --version
+
+      - name: Validate server.json
+        run: mcp-publisher validate server.json
+
+      - name: Publish to MCP Registry (OIDC)
+        run: |
+          mcp-publisher login github-oidc
+          mcp-publisher publish server.json
+        # A registry outage must never break the release; re-run from Actions if needed.
+        continue-on-error: true

dnb_mcp-0.1.0/.github/workflows/smoke.yml

@@ -0,0 +1,21 @@
+name: Daily live smoke
+
+# One fixed probe per day: catches fingerprint rot / bot-wall changes
+# before users hit them. Fails visibly on 403/timeout.
+on:
+  schedule:
+    - cron: "0 6 * * *"
+  workflow_dispatch:
+
+jobs:
+  smoke:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+      - name: Install
+        run: pip install -e .
+      - name: Smoke probe
+        run: python tests/smoke_live.py

dnb_mcp-0.1.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Mohammad Abbasi
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

dnb_mcp-0.1.0/PKG-INFO

@@ -0,0 +1,72 @@
+Metadata-Version: 2.5
+Name: dnb-mcp
+Version: 0.1.0
+Summary: Unofficial read-only MCP server for the Deutsche Nationalbibliothek catalogue (DNB SRU) — books, authors, ISBN
+Author: Mohammad Abbasi
+License: MIT
+License-File: LICENSE
+Requires-Python: >=3.10
+Requires-Dist: httpx>=0.27
+Requires-Dist: mcp<2
+Description-Content-Type: text/markdown
+
+# dnb-mcp 📚
+
+mcp-name: io.github.mohammad-abbasi-sh/dnb-mcp
+
+Unofficial **read-only** MCP server for **[services.dnb.de](https://www.dnb.de)** — Deutsche Nationalbibliothek catalogue via its official SRU API: keyword, title and author search, ISBN and DNB-ID lookup. No key, no login. No key, no login.
+
+> Built by Mohammad Abbasi as part of the "Germany MCP factory". Not affiliated with services.dnb.de.
+
+## For AI agents
+
+Add to your MCP client config (stdio):
+
+```json
+{
+  "mcpServers": {
+    "dnb": {
+      "command": "uvx",
+      "args": ["dnb-mcp==0.1.0"]
+    }
+  }
+}
+```
+
+Or run locally: `python -m dnb_mcp.server` (needs `mcp<2`, `httpx`).
+
+## Tools
+
+| Tool | Description |
+|------|-------------|
+| `dnb_search_books` | Keyword search across the DNB catalogue — `query`, `limit` 1..50 |
+| `dnb_search_by_title` | Title search — `title`, `limit` 1..50 |
+| `dnb_search_by_author` | Author search — `author` (e.g. "Kafka, Franz"), `limit` 1..50 |
+| `dnb_by_isbn` | Look up one book by ISBN — `isbn` (13 or 10 digits, with or without dashes) |
+| `dnb_get_record` | Full record by DNB identifier — `dnb_id` (e.g. "1306459214") |
+| `dnb_health` | Reachability probe (ok / blocked / error) — one polite request |
+
+## Notes
+
+- All data comes from the **Deutsche Nationalbibliothek SRU API** (free, no key, no login).
+- Indexes used (from the API's own explain response): `woe` (all terms), `tit` (title), `per` (person/author), `num` (number incl. ISBN), `idn` (DNB identifier).
+- There is **no newspapers-only filter**: the `mat` (material type) index returns 0 for every German value tried (Bücher, Zeitschriften, Zeitung, Zeitungen, Online-Ressource), so no newspapers-only tool is offered; newspapers are still found by the keyword and title searches.
+- Dublin Core recordSchema is unsupported by the API — MARC21-xml only.
+
+Tests: 20/20 green (14 offline + 6 live, 2026-11-10).
+
+## Development
+
+```bash
+python3 -m venv .venv && .venv/bin/pip install "mcp<2" httpx pytest
+.venv/bin/pip install -e .
+.venv/bin/python -m pytest tests/ -q
+```
+
+## License
+
+MIT
+
+---
+
+Built by [Mohammad Abbasi](https://github.com/mohammad-abbasi-sh) — [Almanibama](https://almanibama.ir) · [mabbasi.ir](https://mabbasi.ir)

dnb_mcp-0.1.0/README-de.md

@@ -0,0 +1,51 @@
+# dnb-mcp 📚
+
+mcp-name: io.github.mohammad-abbasi-sh/dnb-mcp
+
+Inoffizieller **nur lesender** MCP-Server für **[services.dnb.de](https://www.dnb.de)** — Katalog der Deutschen Nationalbibliothek über die offizielle SRU-API: Stichwort-, Titel- und Autorensuche, ISBN- und DNB-ID-Abfrage. Kein Schlüssel, kein Login. Kein Schlüssel, keine Anmeldung.
+
+> Gebaut von Mohammad Abbasi als Teil der "Germany MCP factory". Nicht mit services.dnb.de verbunden.
+
+## Für KI-Agenten
+
+Zur MCP-Client-Konfiguration hinzufügen (stdio):
+
+```json
+{
+  "mcpServers": {
+    "dnb": {
+      "command": "uvx",
+      "args": ["dnb-mcp==0.1.0"]
+    }
+  }
+}
+```
+
+Oder lokal starten: `python -m dnb_mcp.server` (benötigt `mcp<2`, `httpx`).
+
+## Tools
+
+| Tool | Beschreibung |
+|------|-------------|
+| `dnb_search_books` | Stichwortsuche im DNB-Katalog |
+| `dnb_search_by_title` | Titelsuche im DNB-Katalog |
+| `dnb_search_by_author` | Autorensuche im DNB-Katalog |
+| `dnb_by_isbn` | Ein Buch per ISBN nachschlagen |
+| `dnb_get_record` | Vollständiger Datensatz per DNB-ID |
+| `dnb_health` | Erreichbarkeitsprüfung (ok / blocked / error) — eine höfliche Anfrage |
+
+## Entwicklung
+
+```bash
+python3 -m venv .venv && .venv/bin/pip install "mcp<2" httpx pytest
+.venv/bin/pip install -e .
+.venv/bin/python -m pytest tests/ -q
+```
+
+## Lizenz
+
+MIT
+
+---
+
+Built by [Mohammad Abbasi](https://github.com/mohammad-abbasi-sh) — [Almanibama](https://almanibama.ir) · [mabbasi.ir](https://mabbasi.ir)

dnb_mcp-0.1.0/README-fa.md

@@ -0,0 +1,51 @@
+# dnb-mcp 📚
+
+mcp-name: io.github.mohammad-abbasi-sh/dnb-mcp
+
+سرور MCP غیررسمی و **فقط‌خواندنی** برای **[services.dnb.de](https://www.dnb.de)** — کاتالوگ کتابخانه ملی آلمان از طریق API رسمی SRU: جست‌وجوی کلیدواژه، عنوان و نویسنده، جست‌وجو با ISBN و شناسهٔ DNB. بدون کلید، بدون ورود. بدون کلید، بدون ورود.
+
+> ساختهٔ محمد عباسی، بخشی از «کارخانهٔ MCP آلمان». وابسته به services.dnb.de نیست.
+
+## برای ایجنت‌های هوش مصنوعی
+
+به پیکربندی MCP کلاینت اضافه کنید (stdio):
+
+```json
+{
+  "mcpServers": {
+    "dnb": {
+      "command": "uvx",
+      "args": ["dnb-mcp==0.1.0"]
+    }
+  }
+}
+```
+
+یا به‌صورت محلی اجرا کنید: `python -m dnb_mcp.server` (نیازمند `mcp<2` و `httpx`).
+
+## ابزارها
+
+| ابزار | توضیح |
+|------|-------|
+| `dnb_search_books` | جست‌وجوی کلیدواژه در کاتالوگ DNB |
+| `dnb_search_by_title` | جست‌وجو در کاتالوگ DNB با عنوان |
+| `dnb_search_by_author` | جست‌وجو در کاتالوگ DNB با نویسنده |
+| `dnb_by_isbn` | یافتن یک کتاب با ISBN |
+| `dnb_get_record` | رکورد کامل با شناسهٔ DNB |
+| `dnb_health` | بررسی دسترس‌پذیری (ok / blocked / error) — یک درخواست مؤدبانه |
+
+## توسعه
+
+```bash
+python3 -m venv .venv && .venv/bin/pip install "mcp<2" httpx pytest
+.venv/bin/pip install -e .
+.venv/bin/python -m pytest tests/ -q
+```
+
+## لایسنس
+
+MIT
+
+---
+
+Built by [Mohammad Abbasi](https://github.com/mohammad-abbasi-sh) — [Almanibama](https://almanibama.ir) · [mabbasi.ir](https://mabbasi.ir)

dnb_mcp-0.1.0/README.md

@@ -0,0 +1,60 @@
+# dnb-mcp 📚
+
+mcp-name: io.github.mohammad-abbasi-sh/dnb-mcp
+
+Unofficial **read-only** MCP server for **[services.dnb.de](https://www.dnb.de)** — Deutsche Nationalbibliothek catalogue via its official SRU API: keyword, title and author search, ISBN and DNB-ID lookup. No key, no login. No key, no login.
+
+> Built by Mohammad Abbasi as part of the "Germany MCP factory". Not affiliated with services.dnb.de.
+
+## For AI agents
+
+Add to your MCP client config (stdio):
+
+```json
+{
+  "mcpServers": {
+    "dnb": {
+      "command": "uvx",
+      "args": ["dnb-mcp==0.1.0"]
+    }
+  }
+}
+```
+
+Or run locally: `python -m dnb_mcp.server` (needs `mcp<2`, `httpx`).
+
+## Tools
+
+| Tool | Description |
+|------|-------------|
+| `dnb_search_books` | Keyword search across the DNB catalogue — `query`, `limit` 1..50 |
+| `dnb_search_by_title` | Title search — `title`, `limit` 1..50 |
+| `dnb_search_by_author` | Author search — `author` (e.g. "Kafka, Franz"), `limit` 1..50 |
+| `dnb_by_isbn` | Look up one book by ISBN — `isbn` (13 or 10 digits, with or without dashes) |
+| `dnb_get_record` | Full record by DNB identifier — `dnb_id` (e.g. "1306459214") |
+| `dnb_health` | Reachability probe (ok / blocked / error) — one polite request |
+
+## Notes
+
+- All data comes from the **Deutsche Nationalbibliothek SRU API** (free, no key, no login).
+- Indexes used (from the API's own explain response): `woe` (all terms), `tit` (title), `per` (person/author), `num` (number incl. ISBN), `idn` (DNB identifier).
+- There is **no newspapers-only filter**: the `mat` (material type) index returns 0 for every German value tried (Bücher, Zeitschriften, Zeitung, Zeitungen, Online-Ressource), so no newspapers-only tool is offered; newspapers are still found by the keyword and title searches.
+- Dublin Core recordSchema is unsupported by the API — MARC21-xml only.
+
+Tests: 20/20 green (14 offline + 6 live, 2026-11-10).
+
+## Development
+
+```bash
+python3 -m venv .venv && .venv/bin/pip install "mcp<2" httpx pytest
+.venv/bin/pip install -e .
+.venv/bin/python -m pytest tests/ -q
+```
+
+## License
+
+MIT
+
+---
+
+Built by [Mohammad Abbasi](https://github.com/mohammad-abbasi-sh) — [Almanibama](https://almanibama.ir) · [mabbasi.ir](https://mabbasi.ir)

dnb_mcp-0.1.0/pyproject.toml

@@ -0,0 +1,25 @@
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[project]
+name = "dnb-mcp"
+version = "0.1.0"
+description = "Unofficial read-only MCP server for the Deutsche Nationalbibliothek catalogue (DNB SRU) — books, authors, ISBN"
+readme = "README.md"
+license = { text = "MIT" }
+requires-python = ">=3.10"
+dependencies = [
+    "mcp<2",
+    "httpx>=0.27",
+]
+authors = [{ name = "Mohammad Abbasi" }]
+
+[project.scripts]
+dnb-mcp = "dnb_mcp.server:main"
+
+[tool.hatch.build.targets.wheel]
+packages = ["src/dnb_mcp"]
+
+[tool.pytest.ini_options]
+markers = ["live: tests that hit the real https://www.dnb.de site"]

dnb_mcp-0.1.0/server.json

@@ -0,0 +1 @@
+{"$schema": "https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json", "description": "Unofficial read-only MCP server for the Deutsche Nationalbibliothek catalogue (DNB SRU).", "name": "io.github.mohammad-abbasi-sh/dnb-mcp", "packages": [{"identifier": "dnb-mcp", "registryBaseUrl": "https://pypi.org", "registryType": "pypi", "transport": {"type": "stdio"}, "version": "0.1.0"}], "repository": {"source": "github", "url": "https://github.com/mohammad-abbasi-sh/dnb-mcp"}, "title": "DNB — Deutsche Nationalbibliothek catalogue: books, authors & ISBN lookup", "version": "0.1.0", "websiteUrl": "https://github.com/mohammad-abbasi-sh/dnb-mcp"}

dnb_mcp-0.1.0/src/dnb_mcp/__init__.py

@@ -0,0 +1,7 @@
+"""dnb_mcp — unofficial read-only MCP server for services.dnb.de.
+
+Deutsche Nationalbibliothek catalogue via its official SRU API: keyword, title and author search, ISBN and DNB-ID lookup. No key, no login.
+"""
+from __future__ import annotations
+
+__version__ = "0.1.0"

dnb_mcp-0.1.0/src/dnb_mcp/domain.py

@@ -0,0 +1,119 @@
+"""Domain logic: shape DNB SRU (MARC21-xml) responses into compact
+dicts (stdlib only).
+
+Data source: the Deutsche Nationalbibliothek SRU API
+(https://services.dnb.de/sru/dnb) — free, no key, no login.
+CQL indexes used (from the API's own explain response, verified
+2026-11-10): woe (all terms), tit (title), per (person/author),
+num (number — incl. ISBN), idn (DNB identifier).
+
+NOT offered: a newspapers-only search — the `mat` (material type)
+index returns 0 for every German value tried (Bücher, Zeitschriften,
+Zeitung, Zeitungen, Online-Ressource), so no reliable newspaper
+filter exists; newspapers are still found by the keyword/title
+searches. Dublin Core recordSchema is unsupported ("Unknown schema
+for retrieval") — MARC21-xml only.
+"""
+from __future__ import annotations
+
+import re
+import urllib.parse
+import xml.etree.ElementTree as ET
+
+BASE = "https://services.dnb.de/sru/dnb"
+
+SRW = "{http://www.loc.gov/zing/srw/}"
+M = "{http://www.loc.gov/MARC21/slim}"
+
+_ISBN_RE = re.compile(r"^[\d-]{10,17}[0-9Xx]$")
+
+
+def validate_limit(limit: int) -> int:
+    if not 1 <= limit <= 50:
+        raise ValueError(f"limit out of range: {limit} (expected 1..50)")
+    return limit
+
+
+def validate_isbn(isbn: str) -> str:
+    key = (isbn or "").strip().replace(" ", "")
+    if not _ISBN_RE.match(key) or not 10 <= len(key.replace("-", "")) <= 13:
+        raise ValueError(f"invalid ISBN: {isbn!r} "
+                         "(expected 10 or 13 digits, dashes allowed)")
+    return key
+
+
+def validate_dnb_id(dnb_id: str) -> str:
+    key = (dnb_id or "").strip()
+    if not key.isdigit():
+        raise ValueError(f"invalid DNB identifier: {dnb_id!r} "
+                         "(expected digits, e.g. \"1306459214\")")
+    return key
+
+
+def search_url(query: str, maximum_records: int = 10) -> str:
+    params = {
+        "operation": "searchRetrieve",
+        "version": "1.1",
+        "query": query,
+        "recordSchema": "MARC21-xml",
+        "maximumRecords": str(maximum_records),
+    }
+    return BASE + "?" + urllib.parse.urlencode(params)
+
+
+def _subs(record, tag, code):
+    return [s.text.strip() for df in record.findall(f"{M}datafield[@tag='{tag}']")
+            for s in df.findall(f"{M}subfield[@code='{code}']")
+            if (s.text or "").strip()]
+
+
+def _first(record, tag, code):
+    vals = _subs(record, tag, code)
+    return vals[0] if vals else ""
+
+
+def _compact(record) -> dict:
+    title = _first(record, "245", "a")
+    subtitle = _first(record, "245", "b")
+    if subtitle:
+        title = f"{title} : {subtitle}"
+    author = _first(record, "100", "a") or _first(record, "110", "a")
+    return {
+        "dnb_id": (record.findtext(f"{M}controlfield[@tag='001']") or "").strip(),
+        "title": title,
+        "author": author,
+        "contributors": _subs(record, "700", "a")[:3],
+        "year": _first(record, "264", "c"),
+        "publisher": _first(record, "264", "b"),
+        "place": _first(record, "264", "a"),
+        "edition": _first(record, "250", "a"),
+        "isbn": _first(record, "020", "a"),
+        "language": _first(record, "041", "a"),
+        "pages": _first(record, "300", "a"),
+        "subjects": _subs(record, "653", "a")[:5],
+        "url": _first(record, "856", "u"),
+    }
+
+
+def parse_search(text: str) -> dict:
+    """Parse an SRU searchRetrieve response -> {total_results, items}."""
+    try:
+        root = ET.fromstring(text)
+    except ET.ParseError as e:
+        raise ValueError(f"invalid SRU response: {e}") from e
+    diag = root.find(f"{SRW}diagnostics")
+    if diag is not None:
+        details = diag.findtext(".//{http://www.loc.gov/zing/srw/diagnostic/}details") or ""
+        msg = diag.findtext(".//{http://www.loc.gov/zing/srw/diagnostic/}message") or ""
+        raise ValueError(f"DNB SRU error: {details} ({msg})".strip())
+    total = root.findtext(f"{SRW}numberOfRecords")
+    try:
+        total_n = int(total or "0")
+    except ValueError:
+        total_n = 0
+    items = []
+    for rec in root.findall(f"{SRW}records/{SRW}record"):
+        data = rec.find(f"{SRW}recordData/{M}record")
+        if data is not None:
+            items.append(_compact(data))
+    return {"total_results": total_n, "items": items}

dnb_mcp-0.1.0/src/dnb_mcp/http.py

@@ -0,0 +1,99 @@
+"""HTTP layer for dnb_mcp — plain httpx GETs against services.dnb.de.
+
+No key, no login. One retry on 429/5xx; error taxonomy (403/429/other).
+"""
+from __future__ import annotations
+
+import os
+import ssl
+import time
+
+import httpx
+
+UA = (
+    "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 "
+    "(KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36"
+)
+TIMEOUT = 25.0
+PROBE_URL = "https://services.dnb.de/sru/dnb?operation=searchRetrieve&version=1.1&query=woe%3DKafka&recordSchema=MARC21-xml&maximumRecords=1"
+SOURCE = "services.dnb.de"
+
+
+class DnbMcpError(RuntimeError):
+    pass
+
+
+def _verify():
+    """TLS trust store: the OS bundle when present, else httpx/certifi."""
+    for path in (
+        "/etc/ssl/certs/ca-certificates.crt",
+        ssl.get_default_verify_paths().openssl_cafile,
+    ):
+        if path and os.path.exists(path):
+            return ssl.create_default_context(cafile=path)
+    return True
+
+
+def _client() -> httpx.Client:
+    # NOTE: we read the proxy URL from the environment but never log it.
+    proxy = os.environ.get("https_proxy") or os.environ.get("HTTPS_PROXY")
+    return httpx.Client(
+        headers={"User-Agent": UA, "Accept": "application/json, application/xml, text/html, */*"},
+        timeout=TIMEOUT,
+        proxy=proxy or None,
+        trust_env=False,
+        verify=_verify(),
+    )
+
+
+def fetch_text(url: str) -> str:
+    """GET one URL and return its text. One retry on 429/5xx."""
+    last_exc: Exception | None = None
+    r: httpx.Response | None = None
+    for attempt in range(2):
+        try:
+            with _client() as c:
+                r = c.get(url)
+        except httpx.HTTPError as e:
+            last_exc = e
+            if attempt == 0:
+                time.sleep(1.0)
+                continue
+            raise DnbMcpError(f"network error fetching {url}: {e}") from e
+        if r.status_code in (429, 500, 502, 503):
+            if attempt == 0:
+                time.sleep(1.0)
+                continue
+        break
+    assert r is not None  # loop always assigns r or raises
+    if r.status_code == 403:
+        raise DnbMcpError(
+            f"{SOURCE} returned HTTP 403 for {url} (bot wall) — try again later"
+        )
+    if r.status_code == 429:
+        raise DnbMcpError(
+            f"{SOURCE} rate-limited the request (HTTP 429) — try again later"
+        )
+    if r.status_code != 200:
+        raise DnbMcpError(f"{SOURCE} returned HTTP {r.status_code} for {url}")
+    return r.text
+
+
+def health_check() -> dict:
+    """One polite probe of the source site.
+
+    Returns {"status": "ok"|"blocked"|"error", "http_status": int|None,
+    "url": ..., "detail": ...}. Used by the `<prefix>_health` tool
+    and the daily smoke CI.
+    """
+    url = PROBE_URL
+    try:
+        fetch_text(url)
+    except DnbMcpError as e:
+        msg = str(e)
+        if "403" in msg:
+            return {"status": "blocked", "http_status": 403, "url": url, "detail": msg}
+        if "429" in msg:
+            return {"status": "blocked", "http_status": 429, "url": url, "detail": msg}
+        return {"status": "error", "http_status": None, "url": url, "detail": msg}
+    return {"status": "ok", "http_status": 200, "url": url}