Open, read-only MCP server for Irish public data (CSO, Oireachtas, Met Éireann, GeoHive, NTA, PPR, data.gov.ie, Smart Dublin, legislation) for Claude, ChatGPT and Copilot
- ✓Open-source license (MIT)
- ✓Actively maintained (<30d)
- ✓Clear description
- ✓Topics declared
- ✓Documented (README)
claude mcp add ireland-mcp -- npx -y ireland-mcp{
"mcpServers": {
"ireland-mcp": {
"command": "npx",
"args": ["-y", "ireland-mcp"]
}
}
}Resumen de MCP Servers
# Ireland MCP
[](https://github.com/c-mongan/ireland-mcp/actions/workflows/ci.yml)
[](LICENSE)
[](https://modelcontextprotocol.io)
[](#why-its-lean)
**Ireland MCP is a free, read-only MCP server that lets AI assistants answer questions with live Irish public data and citations.**
Hosted endpoint: `https://mcp.irishopendata.ie/mcp` (origin fallback: `https://func-ireland-mcp-aofsjpwgy4hva.azurewebsites.net/mcp`)
Transport: Streamable HTTP. Auth: none. Writes: none.
Website: <https://irishopendata.ie> (canonical). `https://irishopendata.com` and `https://mcp.irishopendata.com/mcp` remain working aliases; see [docs/domain-go-live.md](docs/domain-go-live.md).
Registry: [`io.github.c-mongan/ireland-mcp`](https://registry.modelcontextprotocol.io/v0.1/servers/io.github.c-mongan%2Fireland-mcp/versions/latest) on the official MCP Registry.
<!-- counts:start (generated by `npm run counts`; CI checks it) -->
| Count | Value | Definition |
| --- | ---: | --- |
| Data sources | 29 | Upstream public-data sources, one module each. The `cross` combiner is not a source. |
| Working on the hosted endpoint | 28 of 29 | NTA realtime needs an operator key that the hosted service does not have. |
| Catalogue operations | 74 | 71 source operations plus 3 cross-source ones (`list_sources`, `nearby`, `ireland_snapshot`), all callable through `ireland_call`. |
| Default tools | 7 | What `tools/list` returns: 4 meta tools plus `search`, `fetch` and `nearby`. |
| Tools with `?toolsets=all` | 80 | Every operation as a typed tool, plus the 4 meta tools, `search` and `fetch`. |
| Prompts | 3 | `area_profile`, `commute_check`, `compare_counties`. |
<!-- counts:end -->
## Try it in 60 seconds
1. Add the server to your client. For GitHub Copilot CLI:
`copilot mcp add --transport http ireland https://mcp.irishopendata.ie/mcp`
For VS Code, add `.vscode/mcp.json`: `{ "servers": { "ireland": { "type": "http", "url": "https://mcp.irishopendata.ie/mcp" } } }`.
2. Ask: *"Using the ireland MCP tools, give me the Census 2022 profile of Ennis, Co Clare, and the Met Éireann forecast for Galway. Cite the sources."*
3. Check the answer cites `https://data.cso.ie/table/F1015` and the Met Éireann location-forecast URL. The forecast URL is `http://`; Met Éireann's `https://` host currently redirects to an origin that fails TLS.
<p align="center">
<img src="docs/demo/inspector.gif" alt="MCP Inspector calling the hosted endpoint: nearby Galway, CSO profile for Ennis, Met Éireann forecast and Property Price Register stats, each with a source citation" width="800">
</p>
| Client (tested 7 Oct 2026 against the hosted endpoint) | Result |
| --- | --- |
| MCP Inspector (Playwright, [`scripts/demo-inspector.mjs`](scripts/demo-inspector.mjs)) | 4/4 calls (`nearby`, CSO, Met Éireann, PPR) returned data with citations. [CSO Ennis](docs/demo/inspector-cso-ennis.png), [PPR Galway 2024](docs/demo/inspector-ppr-galway-2024.png). |
| VS Code Insiders, Copilot agent mode | Census 2022 Ennis (27,923 people) and Galway forecast, with CSO and Met Éireann citations. [Screenshot](docs/demo/vscode-copilot-chat.png). |
| GitHub Copilot CLI | 3/3 questions answered with citations: Met Éireann Galway, PPR Cork, CSO Ennis. [Evidence](docs/demo/README.md). |
| Every catalogue operation | <!-- live-ops:start -->[docs/live-all-ops.md](docs/live-all-ops.md) (2026-10-08): 72 of 74 pass, 0 fail; 2 need the NTA key.<!-- live-ops:end --> |
How a call flows:
```mermaid
flowchart TD
Q[Client question] --> T[ireland_call / nearby / search]
T --> V[Schema validation]
V --> C{Cache hit?}
C -- yes --> E[Evidence envelope]
C -- no --> B{Circuit breaker open?}
B -- yes --> S{Stale cache available?}
B -- no --> U[Upstream public API: CSO, Met Éireann, PPR, ...]
U -- ok --> E
U -- 5xx or timeout --> R[One jittered retry]
R -- ok --> E
R -- still failing --> S
U -- other error --> S
S -- yes --> F[Stale cached result: stale=true]
F --> E
S -- no --> X["MCP error: isError=true, UPSTREAM_DOWN with hint; no cited result"]
E --> A[Answer with source URL, licence, attribution, retrieved_at]
```
<p align="center">
<a href="docs/architecture/ireland-mcp-arch-v5.drawio">
<img src="docs/architecture/ireland-mcp-arch-v5.png" alt="Ireland MCP architecture: clients, Azure Functions gateway, observability and public data sources" width="900">
</a>
</p>
## Try asking
- “What is the latest population profile for Galway city or County Cork?”
- “Are there Met Éireann weather warnings near Dublin today?”
- “What did houses sell for in Ennis last year?”
- “When is the next train from Heuston, and are there disruptions?”
- “Who are the current TDs for my constituency, and what bills are active?”
- “What protected habitats, monuments, planning applications or census areas are near this point?”
## See it work
| Proof | Screenshot |
| --- | --- |
| Live landing page, desktop | <img src="docs/screenshots/landing-desktop.png" alt="Ireland MCP live landing page on desktop" width="360"> |
| Live landing page, mobile | <img src="docs/screenshots/landing-mobile.png" alt="Ireland MCP live landing page on mobile" width="180"> |
| Install section with one-click client setup | <img src="docs/screenshots/install-section.png" alt="One-click install section for VS Code, Cursor, Claude, ChatGPT and Copilot CLI" width="360"> |
| Playground populated from a real live MCP call | <img src="docs/screenshots/playground-live-call.png" alt="Playground showing a real cso_search_tables result from the live MCP endpoint" width="360"> |
| MCP Inspector connected to the live endpoint | <img src="docs/screenshots/mcp-inspector-live.png" alt="MCP Inspector showing tools/list and an ireland_call result" width="360"> |
| Real model transcript: question → tool calls → answer | <img src="docs/screenshots/agent-transcript.png" alt="Azure OpenAI agent transcript using Ireland MCP tools" width="360"> |
## Install
| Client | Current setup |
| --- | --- |
| Claude custom connector | Settings → Connectors → **Add custom connector** → URL `https://mcp.irishopendata.ie/mcp` → no auth. |
| ChatGPT developer mode connector | Settings → Apps & Connectors → Advanced settings → enable **Developer mode** → create MCP connector → URL above → no auth. The remote endpoint supports Streamable HTTP; `search`/`fetch` are included for deep-research style clients. |
| VS Code | Install link (paste into a browser or run `open` on it): `vscode:mcp/install?%7B%22name%22%3A%22ireland%22%2C%22type%22%3A%22http%22%2C%22url%22%3A%22https%3A%2F%2Fmcp.irishopendata.ie%2Fmcp%22%7D`. Or add `.vscode/mcp.json`: `{ "servers": { "ireland": { "type": "http", "url": "https://mcp.irishopendata.ie/mcp" } } }`. |
| VS Code Insiders | Install link: `vscode-insiders:mcp/install?%7B%22name%22%3A%22ireland%22%2C%22type%22%3A%22http%22%2C%22url%22%3A%22https%3A%2F%2Fmcp.irishopendata.ie%2Fmcp%22%7D`. |
| Cursor | Deeplink: [`cursor://anysphere.cursor-deeplink/mcp/install?name=ireland&config=...`](cursor://anysphere.cursor-deeplink/mcp/install?name=ireland&config=eyJ0eXBlIjoiaHR0cCIsInVybCI6Imh0dHBzOi8vbWNwLmlyaXNob3BlbmRhdGEuaWUvbWNwIn0=). Decoded config: `{ "type": "http", "url": "https://mcp.irishopendata.ie/mcp" }`. |
| GitHub Copilot CLI | `copilot mcp add --transport http ireland https://mcp.irishopendata.ie/mcp` or put the JSON below in `~/.copilot/mcp-config.json`. |
| Claude Code | `claude mcp add --transport http ireland https://mcp.irishopendata.ie/mcp` |
| Local stdio | npm package target: `npx -y ireland-mcp` after npm publication. Today, run from source: `npm ci && npm run build && node dist/src/cli.js`. To list typed tools locally, run `node dist/src/cli.js --toolsets=cso,irish-rail` or `IRELAND_MCP_TOOLSETS=all node dist/src/cli.js` after building. |
Portable Copilot / MCP JSON:
```json
{
"mcpServers": {
"ireland": {
"type": "http",
"url": "https://mcp.irishopendata.ie/mcp",
"tools": ["*"]
}
}
}
```
## Why it is lean
Most data MCPs expose every typed tool up front. Ireland MCP keeps the default `tools/list` to **7 tools, 8,407 characters, about 2,102 tokens** (`toolsets=all` is 80 tools, about 22k tokens) (`npm run measure:tools`). CI fails if that list grows past 16,000 characters.
<p align="center">
<a href="docs/architecture/ireland-mcp-lean-surface.drawio">
<img src="docs/architecture/ireland-mcp-lean-surface.png" alt="Token-cost comparison: 7 default tools versus all typed tools and one-server-per-source routing" width="800">
</a>
</p>
Need typed tools anyway?
| Surface | Use |
| --- | --- |
| Default HTTP | `/mcp` lists only `ireland_catalogue`, `ireland_describe`, `ireland_call`, `ireland_about`, `search`, `fetch`, `nearby`. |
| Source toolsets | `/mcp?toolsets=cso,irish-rail` lists selected typed tools too. |
| One source path | `/mcp/x/{source}` lists that source's typed tools, for example `/mcp/x/met-eireann`. |
| Everything | `/mcp?toolsets=all` lists all available tools and is useful for debugging, not routine chat. |
| Local stdio | From source: `node dist/src/cli.js --toolsets=cso` or `IRELAND_MCP_TOOLSETS=all node dist/src/cli.js` (npm package not yet published). |
`max_tokens` defaults to about 2,000 output tokens and can be set from 100 to 8,000. Large results return `truncated`, counts and narrowing hints instead of flooding the context.
## Architecture
The server is a stateless Azure Functions Flex app with a lean gateway in front of the source modules. The gateway owns the operation index, response budgets, cache, upstream budgets, circuit breakers, one jittered retry for transient 5xx/timeout failures and evidence envelope; every normal resulLo que la gente pregunta sobre ireland-mcp
¿Qué es c-mongan/ireland-mcp?
+
c-mongan/ireland-mcp es mcp servers para el ecosistema de Claude AI. Open, read-only MCP server for Irish public data (CSO, Oireachtas, Met Éireann, GeoHive, NTA, PPR, data.gov.ie, Smart Dublin, legislation) for Claude, ChatGPT and Copilot Tiene 0 estrellas en GitHub y su última actualización registrada es del 2026-10-09.
¿Cómo se instala ireland-mcp?
+
Puedes instalar ireland-mcp clonando el repositorio (https://github.com/c-mongan/ireland-mcp) o siguiendo las instrucciones del README en GitHub. ClaudeWave también te ofrece bloques de instalación rápida en esta misma página.
¿Es seguro usar c-mongan/ireland-mcp?
+
Nuestro agente de seguridad ha analizado c-mongan/ireland-mcp y le ha asignado un Trust Score de 95/100 (tier: Verified). Revisa el desglose completo de comprobaciones superadas y flags en esta página.
¿Quién mantiene c-mongan/ireland-mcp?
+
c-mongan/ireland-mcp es mantenido por c-mongan. La última actividad registrada en GitHub es del 2026-10-09, con 0 issues abiertos.
¿Hay alternativas a ireland-mcp?
+
Sí. En ClaudeWave puedes explorar mcp servers similares en /categories/mcp, ordenados por popularidad o actividad reciente.
Despliega ireland-mcp en tu cloud
Lleva este repo a producción en minutos. Cada plataforma genera su propio entorno con variables de entorno editables.
¿Mantienes este repo? Añade un badge a tu README
Pega el badge en tu README de GitHub para mostrar que está auditado por ClaudeWave. Cada badge enlaza de vuelta a esta página y muestra el Trust Score actual.
[](https://claudewave.com/repo/c-mongan-ireland-mcp)<a href="https://claudewave.com/repo/c-mongan-ireland-mcp"><img src="https://claudewave.com/api/badge/c-mongan-ireland-mcp" alt="Featured on ClaudeWave: c-mongan/ireland-mcp" width="320" height="64" /></a>Más MCP Servers
Fair-code workflow automation platform with native AI capabilities. Combine visual building with custom code, self-host or cloud, 400+ integrations.
User-friendly AI Interface (Supports Ollama, OpenAI API, ...)
An open-source AI agent that brings the power of Gemini directly into your terminal.
Real-time global intelligence dashboard. AI-powered news aggregation, geopolitical monitoring, and infrastructure tracking in a unified situational awareness interface
🕷️ An adaptive Web Scraping framework that handles everything from a single request to a full-scale crawl! Don't be shy, join here: https://discord.gg/EMgGbDceNQ and follow here for daily tips and tricks: https://x.com/Scrapling_dev
The fastest path to AI-powered full stack observability, even for lean teams.