Unquant

Hosted remote financial data MCP for AI agents, covering market data, company fundamentals, financial news, macroeconomics, and US Congress disclosures.

GitHub

Dokumentation

Unquant — Financial Data MCP for AI Agents

简体中文

Public reference hub · Tool and dataset catalog

Official examples and machine-readable contracts for the hosted Unquant remote Model Context Protocol server. Connect AI agents to market data, company fundamentals, financial news, US macroeconomics, and US Congress disclosures through one Streamable HTTP endpoint:

https://unquant.ai/mcp

No local Unquant package or wrapper is required. This repository contains public connection examples, prompt recipes, and snapshots of contracts already published by the remote server. The production implementation remains private.

28 tool contracts · 22 dataset contracts · 17 currently available

Financial data MCP coverage

Research area	Available coverage	Example dataset IDs
Market and company data	Delayed/end-of-day quotes, price history, symbol search, profiles, fundamentals, earnings, analyst ratings, and ETF holdings	`market.quote`, `market.price_history`, `market.fundamentals`
Financial news	Stock and general news, impact analysis, and article retrieval	`news.stock`, `news.impacts`, `news.article`
Macroeconomics	Economic calendar, indicator history, and current conditions	`macro.calendar`, `macro.indicator`
US Congress disclosures	Normalized transactions and aggregate views	`politics.us_congress_trades`, `politics.us_congress_aggregate`

Availability is explicit in the dataset contract. Planned datasets are documented but are not presented as live.

Public reliability evidence

The scheduled live MCP monitor checks public metadata, OAuth discovery, and tools/list every six hours. Each run publishes HTTP status, request latency, tool-contract consistency, and a machine-readable 90-day artifact. The credential-free monitor never calls a financial dataset. Its public run history is evidence of observed checks, not an unconditional uptime guarantee.

Fastest connection

Claude Code:

claude mcp add --transport http unquant https://unquant.ai/mcp

Generic MCP configuration:

{
  "mcpServers": {
    "unquant": {
      "type": "http",
      "url": "https://unquant.ai/mcp"
    }
  }
}

Other hosts may use a settings UI, command, or different configuration schema. Prefer the current host's native remote-MCP setup.

MCP host type	Setup	Repository evidence
Claude Code	`claude mcp add --transport http unquant https://unquant.ai/mcp`	Connection guide
Remote-MCP clients	Use the native Streamable HTTP URL field	Generic JSON example
Agents inspecting contracts first	Read the tool and dataset snapshots	Machine-readable contracts

Try discovery without installing anything

./examples/curl/list-tools.sh
./examples/curl/list-datasets.sh
./examples/curl/describe-dataset.sh market.quote

The endpoint accepts stateless JSON-RPC tool calls for public discovery. Data tools may use the anonymous Free allowance or OAuth, depending on the current contract and usage state.

Example research recipes

Simplified Chinese versions:

Every financial answer should retain observation dates, reporting periods, units, freshness, warnings, and relevant limitations. Unquant is for research; it does not execute trades or provide personalized investment advice.

Machine-readable contracts

These snapshots describe the current live 2.14.0 contract. A later contract is published here only after the hosted release is live.

Canonical references

Financial data MCP FAQ

What is a financial data MCP server?

A financial data MCP server exposes typed financial-data tools through the Model Context Protocol so an AI agent can discover inputs, call the appropriate dataset, and interpret a structured response. Unquant is a hosted remote MCP server, so the production service does not need to run on the user's computer.

Does Unquant provide real-time stock-market data?

No. Quote and price-history responses are delayed or end-of-day unless a tool response explicitly states otherwise. They are for research, not execution.

Can an AI agent research US Congress trading disclosures?

Yes. The available Congress datasets expose normalized transaction records and aggregate views. See the Congress disclosure recipe.

Where can an agent inspect tools before connecting?

Use the public JSON snapshots above or the endpoint's standard MCP tools/list method. Contract snapshots contain metadata, not returned financial records.

Does Unquant trade or provide investment advice?

No. Unquant provides research data. It does not execute trades or provide personalized investment advice.

Community

Ask setup and interoperability questions in GitHub Discussions.
Request a dataset with the public research outcome and required fields.
Report a reproducible integration problem.
Read CONTRIBUTING.md before proposing examples or documentation changes.

Licensing and data rights

The example code in this repository is MIT licensed. That license does not grant rights to Unquant data or the hosted service. Data access, use, redistribution, and pass-through restrictions are governed by the Unquant Terms and the data-and-service notice.

The JSON files in catalog/ are public contract metadata and documentation under the repository's MIT license. They contain no returned financial records. That documentation license does not grant rights to data returned by the hosted service.

Do not open an issue containing access tokens, email codes, payment details, personal financial data, or licensed third-party content.