@363045841yyt/klinechart-ai-runtime

MCP (Model Context Protocol) server and AI tool schemas for @363045841yyt/klinechart-core (npm) / @363045841yyt/klinechart (npm).

Optional addon — install only if you need AI agent / MCP control of your charts.

Provides a WebSocket-bridged MCP server that enables AI agents (via MCP Inspector or any MCP client) to control K-line chart operations — zoom, pan, add/remove indicators, change theme, and more.

Installation

pnpm add @363045841yyt/klinechart-ai-runtime

Requires @363045841yyt/klinechart-core as a peer dependency.

Quick Start

Start the MCP server

import { createMcpServer } from '@363045841yyt/klinechart-ai-runtime/mcp-server'

const { start, stop } = createMcpServer({
  ws: { port: 8080 },
})

await start()

Integrate with KLineChart (Vue example)

<script setup lang="ts">
import KLineChart from '@363045841yyt/klinechart'
import { executeTool } from '@363045841yyt/klinechart-ai-runtime'

const chartRef = ref<InstanceType<typeof KLineChart> | null>(null)

const mcpConfig = {
  wsUrl: 'ws://localhost:8080',
  autoReconnect: true,
  onToolCall: (call) => {
    const ctrl = chartRef.value?.getController?.()
    if (!ctrl) return { success: false, error: 'Controller not ready' }
    return executeTool(ctrl, call)
  },
}
</script>

<template>
  <KLineChart ref="chartRef" :mcp="mcpConfig" />
</template>

Connect from MCP Inspector

cd packages/ai-runtime
pnpm inspect

Then call tools like chart.zoomToLevel with { "level": 5 }.

Exports

Main entry (@363045841yyt/klinechart-ai-runtime)

| Export | Description | |--------|-------------| | executeTool | Dispatch a tool call to a chart controller | | ALL_TOOLS | Array of all supported tool schemas | | TOOL_GROUPS | Grouped tool definitions | | findTool(name) | Look up a tool schema by name | | describeVolumeProfileState | Generate VP state summary | | describeAnchoredVwap | Generate anchored VWAP summary | | describeFootprintLatestBar | Generate footprint summary | | describeAlerts | Generate alerts summary | | serialize / deserialize | Chart state serialization | | SessionRegistry | WebSocket session manager |

MCP Server (@363045841yyt/klinechart-ai-runtime/mcp-server)

| Export | Description | |--------|-------------| | createMcpServer(options) | Create MCP + WebSocket server instance |

Create with MCP (@363045841yyt/klinechart-ai-runtime/create-with-mcp)

Legacy helper — prefer executeTool + mcp prop pattern above.

Architecture

┌─────────────────┐        WebSocket          ┌───────────────────┐
│   Browser       │◄─────────────────────────►│  MCP Server       │
│                 │   register / tool:call    │  (Node.js)        │
│  KLineChart     │   tool:result /           │                   │
│  └─ ChartBridge │   state:update            │ ┌─ SessionRegistry│
│       ↓         │                           │ └─ WsSessionHandle│
│  └─ onToolCall ─┼───────────────────────────┼──► executeTool    │
└─────────────────┘                           └────────┬──────────┘
                                                       │ stdio
                                                       ▼
                                               ┌────────────────────┐
                                               │  MCP Client        │
                                               │  (Inspector / AI)  │
                                               └────────────────────┘

Available Tools

| Tool | Description | |------|-------------| | chart.zoomToLevel | Zoom to a specific level | | chart.setTheme | Switch between light/dark theme | | indicators.add | Add an indicator by definition ID | | indicators.remove | Remove an indicator by instance ID | | indicators.updateParams | Update indicator parameters |

License

MIT