DEV Community: OpenData Forge

Give Claude/Cursor access to Spain's official gazette (BOE) via MCP

OpenData Forge — Sat, 17 Oct 2026 16:59:20 +0000

Disclosure: I maintain the open-source MCP server used here (MIT licence) and its hosted versions on MCPize (free plan plus paid plans, so I earn from paid subscriptions) and Smithery (free). This article was written with AI assistance, and every command was run and checked before publishing.

Spain's Boletín Oficial del Estado (BOE) publishes new laws, grants, public jobs and tender notices every day, and its sister bulletin BORME publishes company registry acts. Both have an official open-data API, but asking an AI assistant "any SME grants in the BOE this week?" normally gets you a stale or made-up answer.

The Model Context Protocol fixes that by giving the assistant a real tool to call. In about five minutes you'll have Claude Desktop or Cursor querying the live BOE, plus EU funding calls and EU tenders (TED).

The server

eu-spain-public-data-mcp exposes three read-only tools, and none of them need an API key:

Tool	Source
`search_boe_borme`	BOE / BORME daily summaries (keywords, sections, date range up to 31 days)
`search_eu_funding`	EU Funding & Tenders Portal: open and forthcoming calls, sorted by deadline
`search_ted_tenders`	TED procurement notices (keyword, buyer country, CPV, date)

Option A: run it locally (free)

Requires Node 20+.

git clone https://github.com/thehomer87/eu-spain-public-data-mcp
cd eu-spain-public-data-mcp
npm ci && npm run build

Claude Desktop: edit claude_desktop_config.json (Settings → Developer → Edit Config):

{
  "mcpServers": {
    "spain-eu-data": {
      "command": "node",
      "args": ["/absolute/path/eu-spain-public-data-mcp/dist/index.js", "--stdio"]
    }
  }
}

Cursor: put the same block in ~/.cursor/mcp.json (global) or .cursor/mcp.json (per project), then enable it under Settings → MCP.

Restart the app. You should see three tools under "spain-eu-data".

Option B: hosted, nothing to install

The same server runs on MCPize. Subscribe to the free plan (50 calls a month) and copy the connection snippet MCPize shows you into the same config file. Hosting has paid tiers if you need more calls.

Option C: Smithery

The server is also listed on Smithery, which generates the config for Claude Desktop, Cursor and other clients in one click.

Check it works (without an AI client)

Before blaming your config, talk to the server directly over stdio. The repo ships a smoke test that starts the server, lists the tools and makes four real calls:

cd eu-spain-public-data-mcp
npm test

The output from a fresh clone on 17 Oct 2026 (abridged):

stdio tools: search_boe_borme, search_eu_funding, search_ted_tenders
search_boe_borme {"keywords":["subvenci"],"maxResults":5} -> 5 items in 1568ms | first: "Real Decreto 832/2026, de 14 de octubre, por el que se regula la concesión directa de una subvención a la Fun…
search_boe_borme {"bulletins":["BORME"],"maxResults":3} -> 3 items in 257ms
search_eu_funding {"keywords":"artificial intelligence","maxResults":5} -> 5 items in 1038ms | first: "Artificial Intelligence for Cybersecurity applications"
search_ted_tenders {"keywords":"software","countries":["ESP"],"maxResults":5} -> 5 items in 670ms

If that works but your AI client shows no tools, the problem is the config path or the client wasn't restarted. A good first BOE call to try, because the parameter names matter:

{ "keywords": ["ayudas", "subvenciones"], "sections": ["3", "5B"], "dateFrom": "2026-10-12", "dateTo": "2026-10-17", "maxResults": 3 }

Prompts that work well

"Search the BOE from the last 7 days for grants (sections 3 and 5B) mentioning digitalización and summarise who can apply."
"List open EU calls about artificial intelligence closing in the next 90 days, as a table with deadline and programme."
"Find Spanish software tenders (CPV 72000000) published on TED since Monday."
"What did BORME publish yesterday about company formations in Madrid?" (use bulletins: ["BORME"])

BOE section codes cheat sheet

1 general provisions (laws, royal decrees) · 2A/2B appointments and public exams · 3 other provisions (many grant rules) · 5A tender notices · 5B other notices (grant extracts) · BORME A/B company acts.

Gotchas I hit while testing

Keyword matching covers the title and the department name, so "digital" matches everything published by the Ministerio para la Transformación Digital. Combine keywords with sections.
Unknown argument names are silently ignored. from and limit don't work: use dateFrom and maxResults.
A range is capped at 31 days per call. Ask the assistant to page through months.

Data: Agencia Estatal Boletín Oficial del Estado (www.boe.es); © European Union (EU Funding & Tenders Portal, TED), reused under Commission Decision 2011/833/EU. Not affiliated with or endorsed by these bodies.

Monitor EU public tenders (TED) for your keywords with Apify and n8n

OpenData Forge — Sat, 17 Oct 2026 16:58:46 +0000

Disclosure: I built and maintain the TED Actor used here. It's a paid Actor on the Apify Store ($1.50 per 1,000 results), so I earn a share when you run it. This article was written with AI assistance, and every command and workflow was run and checked before publishing.

Every year the EU's Tenders Electronic Daily (TED) publishes hundreds of thousands of procurement notices. If you sell software, consulting or hardware to the public sector, a 9 a.m. message with only the new, still-open tenders matching your keywords saves hours of portal searching.

In this tutorial we build that alert:

An Apify Actor queries the official TED Search API v3 (no scraping, no TED API key).
An n8n workflow runs it every weekday, drops notices you've already seen and anything already closed, and sends you a digest (email, Slack, Telegram, whatever you prefer).

Total cost: the Actor charges $1.50 per 1,000 results, so 50 results every weekday is roughly $1.70 a month, which fits inside the $5 monthly credit on Apify's free plan. Add n8n (free if self-hosted, or an n8n Cloud plan).

1. Try the Actor

Open the EU Public Tenders (TED) Actor on Apify (create a free account if you don't have one) and run it with this input:

{
  "keywords": "ciberseguridad",
  "countries": ["ESP", "PRT"],
  "publishedAfter": "2026-10-01",
  "maxResults": 20
}

Countries are ISO-3 codes. You can also pass cpvCodes, e.g. ["72000000"] for IT services. When I ran this on 17 Oct 2026, it returned 19 notices in about 2 seconds. Each item looks like this:

{
  "publicationNumber": "707641-2026",
  "title": "Spain – IT services: consulting, software development, Internet and support – Servicio Integral de Ciberseguridad SIC27",
  "buyer": "Compras de la Corporación de Radio y Televisión Española S.A.",
  "country": "ESP",
  "publicationDate": "2026-10-14+02:00",
  "deadline": "2026-11-19+01:00",
  "cpv": ["72000000"],
  "noticeType": "cn-standard",
  "url": "https://ted.europa.eu/en/notice/-/detail/707641-2026"
}

Two things I learned from the real output:

noticeType matters. Keyword searches also return contract award notices (can-*), which are useless for bidding. Keep cn-* (contract notices).
Full-text search is broad."ciberseguridad" also matched an EU furniture framework whose text mentions cybersecurity, and buyer-country matches multi-country EU notices. Add CPV codes if you want precision.

2. Get your Apify API token

In Apify Console go to Settings → API & Integrations and copy your personal API token. In n8n, store it as an environment variable APIFY_TOKEN (self-hosted). On n8n Cloud, or if your instance blocks $env access, create a Query Auth credential instead (name token, value = your token) and select it in the HTTP node. Don't hard-code it in the workflow.

3. Import the n8n workflow

Copy the workflow JSON from the appendix at the end of this post and paste it onto an empty n8n canvas (Ctrl/Cmd+V), or save it as a .json file and use Import from File. It has four nodes:

Node	What it does
Every weekday 08:00	Schedule trigger, cron `0 8 * * 1-5`
Run TED Actor	`POST https://api.apify.com/v2/acts/opendata-forge~eu-public-tenders-ted-search/run-sync-get-dataset-items?token=…`, which runs the Actor and returns its dataset in one call
Only new open tenders	Code node: keeps `cn-*` notices whose deadline hasn't passed and that haven't been alerted before (stored in workflow static data)
Build digest	Code node: turns the items into a subject and text body

The Actor ID (opendata-forge~eu-public-tenders-ted-search) is already filled in. The request body uses n8n expressions so it always looks back two days:

{
  "keywords": "ciberseguridad",
  "countries": ["ESP", "PRT"],
  "publishedAfter": "{{ $now.minus({days: 2}).toFormat('yyyy-MM-dd') }}",
  "maxResults": 50
}

The dedupe logic in Only new open tenders :

const seen = $getWorkflowStaticData('global');
seen.ids = seen.ids || [];
const today = new Date().toISOString().slice(0, 10);
const fresh = [];
for (const item of $input.all()) {
  const n = item.json;
  if (!n.noticeType || !n.noticeType.startsWith('cn')) continue; // contract notices only
  if (n.deadline && n.deadline.slice(0, 10) < today) continue; // already closed
  if (seen.ids.includes(n.publicationNumber)) continue; // already alerted
  seen.ids.push(n.publicationNumber);
  fresh.push({ json: n });
}
seen.ids = seen.ids.slice(-2000);
return fresh;

Static data only persists for active (scheduled) workflows, not manual test runs, so expect repeats while you're testing.

4. Add your delivery node

After Build digest , add a Send Email , Slack or Telegram node and map {{ $json.subject }} and {{ $json.text }}. To skip empty days, add an IF node on {{ $json.count }} > 0.

The digest from my test run (5 of 19 notices survived the filter):

• Spain – IT services: consulting, software development, Internet and support – Servicio Integral de Ciberseguridad SIC27
  Compras de la Corporación de Radio y Televisión Española S.A. | deadline: 2026-11-19 | value: see notice
  https://ted.europa.eu/en/notice/-/detail/707641-2026

5. How I tested this

Ran the Actor code locally against the live TED API with the input above (19 notices).
Ran the n8n workflow (n8n 1.123) from the CLI, with the HTTP node pointed at a local mock that returned exactly that dataset. The workflow succeeded and the filter kept 5 open contract notices. The workflow JSON below imports cleanly. I did not call Apify's hosted endpoint in this test, because that needs your token.

Variations

One workflow per client or sector, each with different keywords and CPV codes.
Write rows to Google Sheets or Airtable for a tender pipeline.
Prefer Spanish national tenders? The same pattern works with the BOE/BORME daily search Actor and sections: ["5A"].

Data: TED – Tenders Electronic Daily, © European Union, reused under Commission Decision 2011/833/EU. This tutorial is not affiliated with the Publications Office of the EU.

Appendix: workflow JSON

{
 "name": "TED tender alerts (Apify + n8n)",
 "nodes": [
  {
   "parameters": {
    "rule": {
     "interval": [
      {
       "field": "cronExpression",
       "expression": "0 8 * * 1-5"
      }
     ]
    }
   },
   "name": "Every weekday 08:00",
   "type": "n8n-nodes-base.scheduleTrigger",
   "typeVersion": 1.2,
   "position": [
    0,
    0
   ]
  },
  {
   "parameters": {
    "method": "POST",
    "url": "https://api.apify.com/v2/acts/opendata-forge~eu-public-tenders-ted-search/run-sync-get-dataset-items",
    "sendQuery": true,
    "queryParameters": {
     "parameters": [
      {
       "name": "token",
       "value": "={{ $env.APIFY_TOKEN }}"
      }
     ]
    },
    "sendBody": true,
    "specifyBody": "json",
    "jsonBody": "={\n \"keywords\": \"ciberseguridad\",\n \"countries\": [\"ESP\", \"PRT\"],\n \"publishedAfter\": \"{{ $now.minus({days: 2}).toFormat('yyyy-MM-dd') }}\",\n \"maxResults\": 50\n}",
    "options": {
     "timeout": 300000
    }
   },
   "name": "Run TED Actor",
   "type": "n8n-nodes-base.httpRequest",
   "typeVersion": 4.2,
   "position": [
    260,
    0
   ]
  },
  {
   "parameters": {
    "jsCode": "// Keep only open calls for tender, drop ones already seen in previous runs\nconst seen = $getWorkflowStaticData('global');\nseen.ids = seen.ids || [];\nconst today = new Date().toISOString().slice(0, 10);\nconst fresh = [];\nfor (const item of $input.all()) {\n const n = item.json;\n if (!n.noticeType || !n.noticeType.startsWith('cn')) continue; // contract notices only (no awards)\n if (n.deadline && n.deadline.slice(0, 10) < today) continue; // already closed\n if (seen.ids.includes(n.publicationNumber)) continue; // already alerted\n seen.ids.push(n.publicationNumber);\n fresh.push({ json: n });\n}\nseen.ids = seen.ids.slice(-2000);\nreturn fresh;"
   },
   "name": "Only new open tenders",
   "type": "n8n-nodes-base.code",
   "typeVersion": 2,
   "position": [
    520,
    0
   ]
  },
  {
   "parameters": {
    "jsCode": "const rows = $input.all().map(i => i.json);\nif (!rows.length) return [{ json: { count: 0, text: 'No new tenders today.' } }];\nconst lines = rows.map(n => `• ${n.title}\\n ${n.buyer ?? ''} | deadline: ${n.deadline ? n.deadline.slice(0,10) : 'n/a'} | ${Number.isFinite(Number(n.estimatedValue)) && n.estimatedValue ? Math.round(Number(n.estimatedValue)).toLocaleString('en') + ' EUR' : 'value: see notice'}\\n ${n.url}`);\nreturn [{ json: { count: rows.length, subject: `${rows.length} new EU tenders for your keywords`, text: lines.join('\\n\\n') } }];"
   },
   "name": "Build digest",
   "type": "n8n-nodes-base.code",
   "typeVersion": 2,
   "position": [
    780,
    0
   ]
  }
 ],
 "connections": {
  "Every weekday 08:00": {
   "main": [
    [
     {
      "node": "Run TED Actor",
      "type": "main",
      "index": 0
     }
    ]
   ]
  },
  "Run TED Actor": {
   "main": [
    [
     {
      "node": "Only new open tenders",
      "type": "main",
      "index": 0
     }
    ]
   ]
  },
  "Only new open tenders": {
   "main": [
    [
     {
      "node": "Build digest",
      "type": "main",
      "index": 0
     }
    ]
   ]
  }
 },
 "settings": {},
 "active": false
}