@trendcraft/mcp

A Model Context Protocol server exposing TrendCraft's technical-analysis indicators and signals to MCP clients like Claude Desktop and Cursor over stdio.

The server's differentiator is its structured per-indicator manifest — 96+ entries with whenToUse, pitfalls, synergy, marketRegime, timeframe, and paramHints — that lets an LLM decide which indicator to use, paired with token-aware calc and signal dispatchers built for agentic single-symbol analysis and multi-symbol screening. It ships no data connector: candles are supplied by the caller (paste from a file, or pair with a data-source MCP server).

Install

npm install -g @trendcraft/mcp
# or run on demand
npx @trendcraft/mcp

The package ships a stdio binary, trendcraft-mcp.

Run

Add the server to your MCP client config. For Claude Desktop, edit claude_desktop_config.json:

{
  "mcpServers": {
    "trendcraft": {
      "command": "npx",
      "args": ["-y", "@trendcraft/mcp"]
    }
  }
}

The same mcpServers entry works for Cursor and ~/.claude.json. For Claude Code:

claude mcp add trendcraft -- npx -y @trendcraft/mcp

Smoke-test the binary without a client. MCP requires an initialize handshake before any other request, so send initialize, then the notifications/initialized notification, then tools/list on the same stdio session:

printf '%s\n' \
  '{"jsonrpc":"2.0","id":0,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"smoke","version":"0"}}}' \
  '{"jsonrpc":"2.0","method":"notifications/initialized"}' \
  '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' \
  | npx -y @trendcraft/mcp

The MCP spec requires the initialize exchange before any other request, so send the full sequence above rather than a lone tools/list — the latter is not guaranteed to be answered across protocol versions and transports.

Tools

| Tool | Purpose | |---|---| | list_indicators | Discover indicator kinds, with optional category / regime / timeframe / calcSupported filters. | | get_indicator_manifest | Full manifest for one kind — whenToUse, pitfalls, synergy, paramHints, etc. | | suggest_indicators_for_regime | Indicators suited to trending / ranging / volatile / low-volatility. | | format_manifest_markdown | Render one indicator's manifest as Markdown for prompt embedding. | | calc_indicator | Compute one indicator on caller-supplied candles. lastN trims response size. | | detect_signal | Detect a signal (crossovers, MA alignment, patterns, divergences, squeeze, volume). Returns a sparse firedAt list for cheap screening. | | list_signals | Discover signal kinds supported by detect_signal, with optional shape filter. | | load_candles | Cache candles in the session and return a reusable handle (see below). |

Candle input

calc_indicator and detect_signal accept exactly one candle source: candles (an array of {time, open, high, low, close, volume?} objects — the canonical form), candlesArray (the same rows as [time, open, high, low, close, volume?] tuples, ~40% smaller payload), or candlesRef (an opaque handle returned by load_candles). Use candlesRef when fanning out several tools over the same series so the bars are transmitted only once; a stale or evicted handle returns INVALID_HANDLE, so just call load_candles again.

Documentation

• EXAMPLES.md — end-to-end recipes: regime-driven discovery, multi-symbol screening, series-vs-events signals, synergy, token-budget tuning, error recovery.
• CHANGELOG.md — release history.

Disclaimer

@trendcraft/mcp and the underlying trendcraft library provide technical-analysis primitives for informational and educational purposes only. Outputs — indicator values, signal triggers, and manifest notes — are not investment advice and are not a recommendation to buy, sell, or hold any instrument. You are solely responsible for any trading decisions made using this software.

License

MIT