# Vibe-Trading MCP Server: AI Trading Agent with 36 Tools

> Vibe-Trading is an AI-powered personal trading agent that exposes 36 MCP tools for market research, strategy drafting, and multi-agent trading teams. It solves the problem of integrating advanced trading capabilities into any MCP-compatible client, offering cross-market data and backtesting without initial API keys for core research. Developers building AI assistants for financial analysis will find it useful.

**Published:** 2026-08-19T12:00:24.875+00:00

**Keywords:** vibe-trading,mcp-server,ai-trading-agent,financial-mcp,trading-tools

# Vibe-Trading: Your Personal AI Trading Agent via MCP

Vibe-Trading offers a personal trading agent powered by AI, designed to integrate directly into your existing MCP-compatible clients. It provides a suite of 36 MCP tools, enabling natural-language market research, strategy drafting, and the orchestration of multi-agent trading teams directly from your preferred environment.

## Connecting Vibe-Trading to Your MCP Client

The primary integration path for Vibe-Trading is as an MCP plugin, running as a standard I/O subprocess. This means no complex server setup is required to get started. Core research tools operate without needing API keys for HK/US/crypto markets, making initial exploration straightforward. Trading connector tools, however, utilize a selected connector profile, and `run_swarm` requires an LLM key.

For clients like Claude Desktop, integrate Vibe-Trading by adding a simple configuration entry:

```json
{
  "mcpServers": {
    "vibe-trading": {
      "command": "vibe-trading-mcp"
    }
  }
}
```

OpenClaw users can add Vibe-Trading as a skill in their `config.yaml`:

```yaml
skills:
  - name: vibe-trading
    command: vibe-trading-mcp
```

Before engaging `trading_*` tools, OpenClaw users should perform a research-only smoke test to confirm tool discovery and run a market data or backtest request. This ensures the setup is correct before selecting and checking a trading connector profile.

For other MCP clients like Cursor or Windsurf, Vibe-Trading can be invoked via the command line, defaulting to stdio transport:

```bash
vibe-trading-mcp
```

For web clients requiring SSE, specify the transport:

```bash
vibe-trading-mcp --transport sse
```

## Capabilities Exposed Through MCP

Once connected, Vibe-Trading extends your MCP client with capabilities across several key areas:

*   **Self-Improving Trading Agent**: Leverage natural-language market research, draft strategies, and analyze files or web content. Memory-backed workflows support iterative refinement.
*   **Multi-Agent Trading Teams**: Deploy specialized teams for investment, quant, crypto, and risk analysis. These teams provide streaming progress updates and persisted reports, with workers grounded by fetched market data. The `run_swarm` tool, in particular, orchestrates these teams and requires an LLM key.
*   **Cross-Market Data & Backtesting**: Access data for A/HK/US equities, crypto, futures, and forex. Conduct data fallback and composite backtests, utilizing point-in-time (PIT) data, validation, and run cards.

These features are exposed as 36 distinct MCP tools, allowing fine-grained control and integration into your automated workflows or interactive AI assistant sessions.

## References

*   [Vibe-Trading on GitHub](https://github.com/HKUDS/Vibe-Trading)
*   [Model Context Protocol Documentation](https://modelcontextprotocol.io/introduction)
*   [Vibe-Trading on model-context-protocol.com](https://model-context-protocol.com/servers/)

## Related Repository

- [vibe-trading](https://model-context-protocol.com/servers/vibe-trading)

**Source:** https://model-context-protocol.com/blog/vibe-trading-mcp-server-ai-trading-agent-with-36-tools-mcp-server-guide