An open-source MCP server that quotes the market

What it does TickDB is a hosted real-time market data service covering forex, US/HK/CN stocks, crypto, indices, and commodities. This repository contains the open-source MCP server implementation, Skill configuration files, and client examples that let AI coding assistants and agents query that data through natural language. The MCP server exposes 13 tools, ships with 46 unit tests, and can be self-deployed via Docker, though a hosted endpoint at mcp.tickdb.ai runs the same code.

The interesting bit Most market data vendors sell you a raw firehose and leave the AI integration as an exercise for the reader. TickDB open-sources the plumbing—MCP server and Skill definitions—so Claude, Cursor, and similar clients can pull live tickers without writing REST calls.

Key highlights

• Unified coverage across six markets (US, HK, A-Shares, forex, crypto, indices) via one API key.
• AI-native interfaces: MCP server (13 tools), Skill config, and CLI, alongside standard REST and WebSocket.
• Open-source MCP server with Dockerfile, CI workflows, and 46 unit tests under MIT license.
• Claims 10–50ms end-to-end latency on WebSocket streams; supports tick data, order book depth, K-lines, and recent trades.
• Free tier available without a credit card; hosted MCP endpoint mirrors the open-source mcp/ directory.

Caveats

• The repository is primarily interface glue—MCP server, skill configs, and client examples—rather than the underlying data aggregation or exchange connectivity backend, which remains a proprietary hosted service.
• The 10–50ms latency figure appears on a badge but lacks published methodology or reproducible benchmarks in the README.

Verdict Worth a look if you are building AI agents, quant tools, or trading dashboards and want to outsource market data plumbing. Skip it if you need a fully self-hosted data backend or are allergic to hosted API keys.