dsh-stockmcp

English | 简体中文 | 日本語

Stock market intelligence for DeepSeek Harness (DSH) — quotes, financial statements, SEC filings, earnings and economic data, dynamically synced from the stockmcp.net remote MCP server as model tools.

The tool list and parameter schemas are never frozen: they are re-synced at activation, every 10 minutes, and on demand whenever a call reports an unknown tool.

Features

• 📈 Real-time quotes & price history, biggest gainers / losers / most actives
• 📊 Income statements, balance sheets, cash flow, dividends, splits
• 🗂️ SEC filings (list / sections / insider trades), company news
• 🌍 Economic indicators, treasury rates, sector & industry snapshots
• ⚙️ Hot-reloaded API key: settings page, settings.yaml, or environment variable
• 🧩 Zero dependencies — pure ESM, no build step

Requirements

• DeepSeek Harness (dsh CLI) with a web profile
• Node.js >= 20

Install (one command)

dsh plugin --profile web add dsh-stockmcp

or use the one-key installer (it handles pnpm 11 quirks like the 24-hour minimum release age for brand-new versions):

bash  Migrating from the old manual mount? Remove any `./node_modules/StockMCP/...` insert from your profile's `cordis.patch.yml` before restarting, to avoid double-mounting.

## API key

Tool calls require an API key from stockmcp.net. Configure it any of three ways:

1. **Settings page** — DSH web → Settings → StockMCP (writes to `~/.dsh/settings.yaml`, hot-reloaded)
2. **Config file** — add `stockmcp: { apiKey: ... }` to `~/.dsh/settings.yaml` (hot-reloaded)
3. **Environment variable** — `STOCKMCP_API_KEY`

Without a key the plugin stays up, but tool calls return a clear configuration error.

## How it works

The host plugin wraps the stockmcp.net MCP server (JSON-RPC 2.0 over HTTPS) as model tools and exposes same-origin `/stockmcp-api` routes for the settings page (`lib/client.js`).

## Development & publishing

```bash
pnpm pack --dry-run   # inspect what would be published
pnpm publish          # publish to npm (run pnpm login first)

License

MIT