mcp-i14y

An MCP server that connects AI assistants to the [Swiss I14Y Interoperability Platform](https://www.i14y.admin.ch).

Ask Claude, Gemini, or GitHub Copilot to explore Swiss government datasets, APIs, codelists, and public services in natural language — powered by the Model Context Protocol.

Inspired by the French datagouv-mcp project, adapted for the Swiss I14Y platform.

What is I14Y?

The I14Y Interoperability Platform (operated by the Swiss Federal Statistical Office) is Switzerland's national catalog for government data. It standardises and publishes metadata for:

• Datasets — data published by federal and cantonal bodies
• Data Services — APIs exposing government data
• Concepts — reusable data definitions, codelists, and data dictionaries
• Public Services — administrative services offered to citizens and businesses

Features

Quick Start

Option 1: Docker Compose (recommended)

git clone https://github.com/fgouzi/mcp-i14y.git
cd mcp-i14y
cp .env.example .env
docker compose up -d

The server will be available at http://localhost:8400/mcp.

Option 2: Local setup with uv

git clone https://github.com/fgouzi/mcp-i14y.git
cd mcp-i14y
uv sync
uv run python main.py

Verify

curl http://localhost:8400/health
# → {"status":"ok","platform":"i14y","version":"0.1.0"}

Connect to your AI assistant

The MCP endpoint is http://localhost:8400/mcp (Streamable HTTP / JSON-RPC).

Claude Desktop

Add to claude_desktop_config.json:

macOS/Linux: ~/.claude/claude_desktop_config.json · Windows: %APPDATA%\Claude\claude_desktop_config.json

{
  "mcpServers": {
    "i14y": {
      "type": "http",
      "url": "http://localhost:8400/mcp"
    }
  }
}

Claude Code (CLI)

/mcp add i14y http://localhost:8400/mcp

GitHub Copilot (VS Code)

Requires VS Code 1.112+ with the GitHub Copilot Chat extension.

A .vscode/mcp.json file is already included in this repository. After starting the server:

1. Open the repo in VS Code — a "Start" button appears at the top of .vscode/mcp.json
2. Click Start (or run MCP: List Servers from the Command Palette)
3. Open Copilot Chat (Ctrl+Shift+I / Cmd+Shift+I) and switch to Agent mode
4. Click the Tools icon to confirm the i14y tools are available

To configure manually in another project, create .vscode/mcp.json:

{
  "servers": {
    "i14y": {
      "type": "http",
      "url": "http://localhost:8400/mcp"
    }
  }
}

Mistral Le Chat

1. Open Le Chat → side panel → Intelligence → Connectors
2. Click + Add Connector → Custom MCP Connector
3. Fill in:

• Connector name: i14y
• Connection server URL: http://localhost:8400/mcp
• Authentication: None

1. Click Connect
2. Enable it in any conversation via the Tools icon (four squares) below the chat input

Mistral Vibe (CLI agent)

Add to your config.toml:

[[mcp_servers]]
name = "i14y"
transport = "streamable-http"
url = "http://localhost:8400/mcp"

Tools are then available as i14y_list_datasets, i14y_get_concept, etc.

MCP Inspector (testing)

npx @modelcontextprotocol/inspector http://localhost:8400/mcp

Available Tools

Datasets

#### list_datasets List datasets from the I14Y platform with optional filters.

#### get_dataset

#### get_dataset_structure

Data Services

#### list_dataservices Same filter parameters as list_datasets (with dataservice_identifier instead of dataset_identifier).

#### get_dataservice

Concepts

#### list_concepts

#### get_concept

#### get_concept_codelist

Public Services

#### list_publicservices Same filter parameters as list_datasets (with publicservice_identifier).

#### get_publicservice

Catalogs

#### get_catalog

Mapping Tables

A mapping table defines a correspondence between two codelists (source → target), enabling semantic alignment across classification systems (e.g. old canton codes → new codes, Swiss codes → European standards).

#### list_mappingtables

#### get_mappingtable

#### get_mappingtable_relations

Search

Since the I14Y API has no full-text search endpoint, these tools fetch pages progressively and rank results client-side.

#### catalog_search Full-text search across all I14Y resource types (server-side, via CORE API).

Concepts (CORE)

#### get_concept_by_identifier

#### get_codelist_entries

#### get_codelist_entry_by_code

#### get_codelist_entries_children

#### search_codelist_entries

Datasets (CORE)

#### get_dataset_by_identifier

#### check_dataset_has_structure

Returns true if a structural model exists, false otherwise. Use this before calling get_dataset_structure() to filter datasets that have a documented schema.

#### get_dataset_model_graph

Returns the dataset schema as a graph (nodes/edges) — suited for programmatic processing. Complements get_dataset_structure() which returns RDF/JSON-LD.

Data Services (CORE)

#### get_dataservice_by_identifier

Public Services (CORE)

#### get_publicservice_by_identifier

#### get_publicservice_relations

Catalogs (CORE)

#### list_catalogs

#### get_catalog_records

#### get_catalog_themes

Agents

#### list_agents

#### get_agent

Vocabularies

Controlled vocabularies define valid values for DCAT-AP metadata fields (themes, access rights, licenses, media types, etc.).

#### list_vocabularies No parameters. Returns all available vocabulary configurations with their identifiers.

#### get_vocabulary

Distributions

#### get_distribution_content Fetch the actual data content of a DCAT distribution file.

Typical workflow:

1. Call get_dataset(dataset_id) → find distributions[n].downloadUrl.uri
2. Call get_distribution_content(url) → returns the file content

Supported formats: JSON, CSV, XML, RDF/Turtle, GeoJSON, plain text. Binary formats (PDF, ZIP, Excel, images) are rejected with an error message.

Example Prompts

Once connected, try asking your LLM assistant:

• "List all public datasets published by the Federal Statistical Office."
• "What data services are available with status Standard?"
• "Show me the codelist for Swiss canton codes."
• "Get the full details for dataset with ID `abc-123`."
• "Export the catalog of the OFS in Turtle format."
• "What public administrative services are registered on I14Y?"
• "Search for datasets about employment by canton."
• "Find the concept codelist for 'Secteur économique'."
• "List all mapping tables that link old canton codes to new ones."
• "Export the mapping relations for mapping table `xyz-456` as CSV."
• "Find the dataset with identifier `px-x-0602010000_109`."
• "Get the valid values for DCAT themes (data subjects)."
• "Which organisations publish data on I14Y?"
• "Get all child codes of canton 1 in the commune hierarchy."

Development

Setup

git clone https://github.com/fgouzi/mcp-i14y.git
cd mcp-i14y
uv sync
cp .env.example .env

Run tests

# Unit tests only (no network required)
uv run pytest

# Including integration tests (hits live I14Y test API)
uv run pytest -m integration

Lint & format

uv run ruff check --fix .
uv run ruff format .

Install pre-commit hooks

uv run pre-commit install

Configuration

Copy .env.example to .env and adjust:

License

MIT — see LICENSE.

Credits

• Inspired by datagouv-mcp by the data.gouv.fr team
• Built on FastMCP and the MCP Python SDK
• Data provided by the Swiss Federal Statistical Office (OFS/BFS) via I14Y