# nordnet-mcp

A [Model Context Protocol](https://modelcontextprotocol.io) server exposing
Nordnet's **public** market data — quotes, charts, news, trade tape, fund/ETF/
stock profiles and FX rates — via the [`nordnet-sdk`](https://pypi.org/project/nordnet-sdk).

**No account. No API key. No login.** This talks to the same API Nordnet's
website uses for anonymous visitors, so it runs anywhere, needs zero
configuration, and holds no secrets.

## Install

```bash
pip install nordnet-mcp
# or with pipx for an isolated runtime:
pipx install nordnet-mcp
```

Requires Python 3.10+.

## Run

```bash
nordnet-mcp            # console script
python -m nordnet_mcp  # or module form
```

The server speaks MCP over **stdio** — wire it into any MCP client.

## Configuration

Configuration is read from environment variables; defaults target Nordnet's
Finnish site and the public market-data host:

| Variable | Default | Purpose |
| --- | --- | --- |
| `NORDNET_BASE_URL` | `https://www.nordnet.fi` | main site host (`.se`/`.dk`/`.no` also work) |
| `NORDNET_MARKET_DATA_BASE_URL` | `https://api.prod.nntech.io` | charts, indices, screening host |
| `NORDNET_LOCALE` | `fi-FI` | sent as `x-locale` on the market-data host |

## Client wiring

Claude Desktop (`claude_desktop_config.json`):

```json
{
  "mcpServers": {
    "nordnet": {
      "command": "nordnet-mcp"
    }
  }
}
```

Hermes Agent (`~/.hermes/config.yaml`):

```yaml
mcp_servers:
  nordnet:
    command: /path/to/nordnet-mcp
```

Generic stdio MCP client:

```json
{ "mcpServers": { "nordnet": { "command": "uvx", "args": ["nordnet-mcp"] } } }
```

## Tools

| Tool | Returns |
| --- | --- |
| `search(query, limit)` | free-text search grouped by asset class |
| `instrument(instrument_id)` | full snapshot: identity, quote, returns, key ratios |
| `quote(instrument_id)` | latest price block with identity context |
| `trades(instrument_id, count)` | recent executed trades (the tape) |
| `stocklist(limit, sort_attribute, sort_order, risers_only)` | market overview (risers/fallers) |
| `fund(instrument_id)` | fund snapshot: NAV, fees, risk class, SFDR article |
| `fund_profile(instrument_id)` / `fund_profile_by_slug(slug)` | full fund screening profile |
| `etf_profile(instrument_id)` / `etf_profile_by_slug(slug)` | full ETF screening profile |
| `stock_profile(instrument_id)` / `stock_profile_by_slug(slug)` | full stock screening profile |
| `chart(instrument_id, period)` / `chart_by_identifier(identifier, period)` | price time series (8 periods) |
| `news(instrument_id, limit, offset)` | paginated news feed |
| `indicators(list_id)` | index/FX quotes for a market list |
| `indicator(slug)` | single indicator / FX quote (e.g. `eurusd`) |
| `market_hours(instrument_id)` / `market_hours_for_order_book(order_book_id)` | trading sessions; `null` when none |

Chart periods: `DAY_1`, `WEEK_1`, `MONTH_1`, `MONTH_3`, `MONTH_6`, `YTD`,
`YEAR_1`, `ALL`.

Instrument IDs are Nordnet's internal numeric IDs (e.g. `16100884` = Wärtsilä
B, `17794251` = Nordnet One Rohkea EUR). Find them with `search`, or read them
from instrument pages on nordnet.fi.

## Public data boundary

Everything served here is data Nordnet shows to anonymous visitors. The one
notable exception: full order book depth (the multi-level bid/ask ladder) is
login-only — anonymously you only get top-of-book (best bid/ask with resting
volumes). This server respects that boundary by design.

## Development

```bash
pip install -e ".[dev]"
pytest           # offline — runs against a local fixture server
uv build         # wheel + sdist
```

## Legal

Unofficial, not affiliated with or endorsed by Nordnet. Only reads public
data; please be polite to the API (the SDK retries transient failures with
backoff). Use at your own risk; market data may be delayed. MIT licensed.