npm.io
0.6.0 • Published 2d agoCLI

@llmquant/data-mcp

Licence
MIT
Version
0.6.0
Deps
2
Size
1.2 MB
Vulns
0
Weekly
0
SummaryDependencyVersions

LLMQuant Data

@llmquant/data-mcp

The knowledge harness for AI‑native finance.
Website · Docs · 中文

npm version npm downloads license node version

Context engineering meets financial data — structured for agent context, not human browsing.

Table of Contents

What It Does

This is an MCP server for LLMQuant Data. It connects any AI agent to financial data — wiki articles, research papers, crypto & equity prices, prediction markets, macro indicators, SEC filings, and more — through the Model Context Protocol.

Configure once — every agent in your stack gets the data.

Available in every agent harness

We're also building llmquantdata-skills — a companion set of agent skills that combine these data tools into ready‑made AI‑native financial workflows (equity research, macro analysis, etc.). Stay tuned.

Available Tools

The credit model is in beta — amounts below may change. Sign up at llmquantdata.com for free credits, no credit card required.

Tool Description Credit
wiki_search Semantic search over 50,000+ quant wiki entries 1
wiki_read Read a wiki article by ID 0
paper_search Semantic search over 1,200+ research papers 1
paper_read Read paper sections (intro, methods, conclusion, …) 0
crypto_historical_klines Crypto OHLCV candlestick data 1
crypto_snapshot Current crypto price + 24h stats 0
polymarket_event_browse List or exact-filter finance-scoped Prediction Markets events by keyword, status, tag, asset, volume, and liquidity filters 1
polymarket_event_search Semantic search over finance-scoped Prediction Markets events 2
polymarket_event_read Read one Prediction Markets event card with child market previews 0
polymarket_market_read Read one Prediction Markets market card with outcomes and outcome token ids 0
polymarket_price_history Hourly or daily implied-probability history for one outcome token 0
equity_historical_prices US equity daily OHLCV + dividend/split data 0
equity_intraday_prices US equity 1h regular-session OHLCV bars 1
macro_indicator_search Browse 50+ curated macro indicators 0
macro_indicator_history Historical observations for a macro indicator 1
macro_indicator_snapshot Latest value for a macro indicator 0
sec_filing_browse Browse SEC 10-K / 10-Q / 8-K filing metadata (each row carries sectionKeys = available section codes) 0
sec_filing_read Read section text from a SEC filing; pass items to fetch a batch in one call, omit for all (8-K requires accession_number — browse first) 1
sec_13f_list_manager_holdings List an institutional manager's 13F holdings (covered manager set × at least the last 4 quarters) 1
sec_13f_list_ticker_holders List institutional holders of a ticker (covered manager set × at least the last 4 quarters) 1
sec_13f_list_top_managers List the top N smart money managers ranked by 13F reportable value (latest quarter, up to 1000) 0
etf_lookup ETF fund identity + SEC mapping + top holdings summary (currently supported ETFs) 0
etf_holdings ETF latest available SEC N-PORT regulatory holdings (currently supported ETFs) 1*
personal_holdings Read the holdings you saved in your LLMQuant Dashboard → Profile (your own account only) 0
personal_profile Read the financial profile you saved in your LLMQuant Dashboard → Profile (your own account only) 0

* An unsupported ETF ticker returns 200 with an explicit coverage notice and 0 credits — you are only charged the 1 credit when holdings are actually returned.

Prediction Markets tools follow an event-first flow: use polymarket_event_search first for natural-language discovery, use polymarket_event_browse only for list/exact-filter requests, then read a selected event, read a market, and request price history for an outcome token.

More data products (company fundamentals, earnings transcripts, etc.) are on the roadmap.

Four ways to access each data product:

Four data access patterns — source data, semantic search, PageIndex tree, knowledge graph

Try It Out

Launch the MCP Inspector to explore tools interactively in your browser:

export LLMQUANT_API_KEY=your_api_key
npx @modelcontextprotocol/inspector npx -y @llmquant/data-mcp

To persist the key, add export LLMQUANT_API_KEY=your_api_key to your ~/.zshrc or ~/.bashrc.

MCP Inspector — interactively test tools in your browser

You'll need an API key — sign up free at llmquantdata.com.

Quick Install (Prompt)

The fastest way: drop this prompt into Claude Code, Cursor, Codex CLI, Gemini CLI, or any MCP-enabled agent. The agent will read this README and run the right setup for its runtime.

Install the LLMQuant data-mcp server in this environment by following https://github.com/LLMQuant/data-mcp

Make sure LLMQUANT_API_KEY is exported in the shell where the agent runs. Get a key at llmquantdata.com.

Prefer to set it up by hand? See Client Setup below.

Client Setup

Claude Code
claude mcp add llmquant-data \
  -e LLMQUANT_API_KEY=your_api_key \
  -- npx -y @llmquant/data-mcp
Cursor

Add to .cursor/mcp.json (project) or ~/.cursor/mcp.json (global):

{
  "mcpServers": {
    "llmquant-data": {
      "command": "npx",
      "args": ["-y", "@llmquant/data-mcp"],
      "env": {
        "LLMQUANT_API_KEY": "your_api_key"
      }
    }
  }
}
Codex CLI
codex mcp add llmquant-data \
  --env LLMQUANT_API_KEY=your_api_key \
  -- npx -y @llmquant/data-mcp
Gemini CLI
gemini mcp add -s user \
  -e LLMQUANT_API_KEY=your_api_key \
  llmquant-data \
  npx -y @llmquant/data-mcp
Other MCP Clients

Any client supporting stdio transport can use this JSON config:

{
  "mcpServers": {
    "llmquant-data": {
      "command": "npx",
      "args": ["-y", "@llmquant/data-mcp"],
      "env": {
        "LLMQUANT_API_KEY": "your_api_key"
      }
    }
  }
}

We're working on integration guides for more clients. If your agent framework isn't listed, open an issue and we'll add it.

Remote / Hosted MCP

@llmquant/data-mcp keeps local stdio as the default setup path. Signed-in LLMQuant users can also generate a hosted Remote MCP connector URL from the dashboard. Hosted connectors are useful when your MCP client supports remote connectors and you do not want to run a local Node.js process.

Claude custom connectors

In the LLMQuant dashboard, open Connect and generate a Remote MCP URL.

https://mcp.llmquantdata.com/u/lqd_mcp_.../mcp

Paste the full URL into Claude's custom connector flow. The hosted URL is separate from your local LLMQUANT_API_KEY and can be revoked from the dashboard.

Treat the hosted connector URL like a password. Anyone with the URL can use the connector until you revoke it.

Environment Variables

Variable Required Default Description
LLMQUANT_API_KEY Yes for stdio — User API key for local stdio mode
LLMQUANT_BASE_URL No https://api.llmquantdata.com API base URL
LLMQUANT_API_TIMEOUT_MS No 15000 Request timeout in ms (max 120000)

Roadmap

  • Streamable HTTP transport (remote MCP without local Node.js)
  • OAuth connector flow for Claude Connectors Directory
  • More data products — company fundamentals, earnings transcripts
  • Agent skills companion package (llmquantdata-skills)
  • Integration guides for more agent frameworks

Have a feature request? Open an issue or email us at contact@llmquant.com.

See Also

  • Awesome Trading Agents — community-curated agent recipes that turn data-mcp tools into end-to-end research workflows (equity research, macro briefings, smart-money digests).
  • Documentation — full API reference, MCP setup guides, and credit policy.

Contributing

This repository is a read-only mirror. We do not accept pull requests.

Found a bug or have a feature request? Please open an issue — we actively triage and respond.

License

MIT