AI & Agents

Twelve Data MCP Server

GitHub: https://github.com/twelvedata/mcp

The Twelve Data MCP Server connects Claude, ChatGPT, and other AI assistants that support the Model Context Protocol (MCP) to Twelve Data’s real-time and historical market data: prices, 60+ technical indicators, fundamentals, ETFs and funds, company news, market movers, and regulatory filings.

There are two ways to use it:

• Hosted server (recommended): https://mcp.twelvedata.com/mcp (streamable HTTP). Add it as a custom connector in your assistant and log in with your Twelve Data account when prompted — nothing to install and no API key to copy. Authorization is OAuth 2.1; client registration and discovery are described in https://twelvedata.com/auth.md.
• Run locally: install the open-source server from the repository and set TWELVE_DATA_API_KEY. It runs over stdio, for example with Claude Desktop, and every request uses your API key.

Typical use cases:

• Connecting Twelve Data to AI assistants or desktop clients that support MCP
• Automating market data access for analysis, dashboards, or bots
• Enabling developers to use Twelve Data as a unified backend for financial data

For supported clients, see the official list here: https://modelcontextprotocol.io/clients

For source code and installation instructions, visit the repository: https://github.com/twelvedata/mcp

Twelve Data on platforms you already use.

• ChatGPT — the Twelve Data plugin in the ChatGPT plugin directory
• OpenClaw — the official Twelve Data skill for OpenClaw and ClawHub
• NEAR Agent — the Twelve Data agent on NEAR's agent.market

Twelve Data is available as a plugin in the ChatGPT plugin directory. Install it once, then mention @Twelve Data in any conversation — there is no separate chat to open. It gives traders, investors, analysts, and developers natural language access to the Twelve Data API without leaving ChatGPT.

Key capabilities:

• Quote snapshots with the current price and today's change
• Real-time and historical price action drawn as an interactive chart inside the conversation
• Technical indicators on demand, such as the RSI for a given symbol
• Search across the Twelve Data endpoint catalog
• Market data workflows for stocks, ETFs, funds, forex, crypto, and more

The plugin is read-only: it looks up and displays market data, and never places trades or moves money.

Example prompts:

• @Twelve Data Get the latest quote and today's change for a stock AAPL
• @Twelve Data Show daily time series for AAPL for the last 30 days
• @Twelve Data What is the RSI for TSLA?

👉 Install it here: Twelve Data in the ChatGPT plugin directory

Twelve Data Official is the official Twelve Data skill for OpenClaw and ClawHub. It provides real-time and historical financial market data for stocks, forex, crypto, ETFs, indices, and more — including quotes, time series, 100+ technical indicators, fundamentals, symbol search, and portfolio/research workflows.

Install the skill:

openclaw skills install @twelvedata/twelvedata

The skill connects to the Twelve Data API through the MCP server (recommended):

• Get an API key at https://twelvedata.com/register
• Run via UV: uvx mcp-server-twelve-data --apikey YOUR_TWELVE_DATA_API_KEY
• Or via pip: pip install mcp-server-twelve-data and mcp-server-twelve-data --apikey YOUR_TWELVE_DATA_API_KEY
• Configure your AI client to use the twelvedata MCP server

When to activate the skill:

• Current prices, quotes, and market snapshots
• Historical time series and charts
• Technical indicators (RSI, MACD, SMA, Bollinger Bands, and 100+ more)
• Fundamentals, earnings, and dividends
• Symbol search and instrument reference data
• Portfolio tracking and research workflows across stocks, forex, crypto, ETFs, and indices

👉 Get the skill here: Twelve Data Official on ClawHub

The official Twelve Data agent is listed on agent.market, NEAR's marketplace for hiring AI agents. It provides premium real-time and historical financial market data for stocks, forex, crypto, and ETFs directly to NEAR-based agents and workflows.

The agent supports both structured queries and natural language queries via AI routing, making it a good fit for:

• Trading signals
• Portfolio agents
• Research
• Backtesting

Jobs are priced in USDC through the marketplace, held in escrow until the result is delivered.

👉 Access it here: Twelve Data on agent.market

Twelve Data is designed with modern AI workflows in mind. For Large Language Models (LLMs) and developers building custom agents, autonomous tools, RAG pipelines, or LLM-powered financial assistants on top of the Twelve Data API, we provide our complete API documentation in machine-readable Markdown format.

https://twelvedata.com/docs/llms.txt

This format is optimized for programmatic ingestion and enables straightforward parsing, indexing, and embedding into LLM pipelines.

The documentation includes:

• Full coverage of all available endpoints
• Detailed parameter definitions and constraints
• Request and response schemas
• Example payloads and usage patterns

This allows you to reliably generate tools, functions, and autonomous agents that interact with the Twelve Data API with minimal manual preprocessing.

The Twelve Data World Model Dataset is an open, multi-modal financial dataset for LLM fine-tuning, time-series foundation models, and world-model / predictive-dynamics training.

• GitHub (pipeline source): https://github.com/twelvedata/twelvedata-world-model-dataset
• Hugging Face (published dataset): https://huggingface.co/datasets/twelvedata/financial-world-model

Coverage and schema:

• 51 large-cap US equities across sectors, with macro context tickers (SPY, QQQ, VIXY, TLT, sector SPDRs) joined onto every equity row
• Multiple timeframes, with the full history available from Twelve Data per symbol
• Three modalities per timeframe:
  • bars_{tf} — OHLCV plus causal technical indicators and macro context (Parquet, 33+ columns)
  • text_{tf} — instruction-tuning text derived from the bars (JSONL)
  • trajectories_{tf} — rolling windows with states / next_states for world-model and RL training (Parquet)

How to load:

from datasets import load_dataset

bars = load_dataset("twelvedata/financial-world-model", "bars_1day")
text = load_dataset("twelvedata/financial-world-model", "text_1h")
traj = load_dataset("twelvedata/financial-world-model", "trajectories_1day")

The Parquet files can also be queried directly with DuckDB.

The dataset is released under the MIT license — see the repository for licensing and data-redistribution notes.