Unquant financial-data MCP examples

August 3, 2026 · View on GitHub

Unquant is a hosted financial-data MCP server for AI agents. It covers market data, company fundamentals, financial news, macroeconomic data, US Congress disclosures, screening, signals, backtesting, and portfolio research.

The production endpoint is:

https://unquant.ai/mcp

The production code and provider credentials are not in this repository.

Paste this into your agent

Connect to https://unquant.ai/mcp. Search the Unquant catalog for the data or
research capability I request. Use catalog.list with a plain-language query,
then use catalog.describe for the best matches. Tell me whether each match is
available, in private preview, or capacity limited. If it is available, use its
listed tool. If it is not available, explain the next action shown by the
catalog. Do not claim access that the catalog does not show.

The agent can answer whether a dataset or research capability exists, is available now, or has another access state. A person only needs to check that the endpoint is unquant.ai and give the request to an agent.

What the agent does

It connects to the remote MCP endpoint.
It searches the broad catalog with catalog.list.
It checks the best matches with catalog.describe.
It uses a listed live tool when access is available.
It reports the catalog state and next action when access is not available.

For the smaller operational dataset catalog, the agent can also use datasets.search and datasets.describe.

Connect

Claude Code:

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

For another client, use its native remote-MCP settings with Streamable HTTP. See the generic configuration and Claude Code guide.

Public contracts

These snapshots are generated from the public production JSON endpoints. The sync job can change only the four JSON files listed above. It cannot copy private source code, environment files, credentials, or returned financial records into this repository.

Use the live MCP endpoint at runtime. The snapshots help search engines, answer engines, and agents discover the service before they connect.

Examples

Limits and support

Unquant is for research. It does not execute trades or provide personalized investment advice. Keep dates, reporting periods, units, warnings, and pagination metadata when you use results.

Use GitHub Discussions for setup questions. Report security problems privately as described in SECURITY.md.

Example code is MIT licensed. Use of the hosted service and returned data is governed by the Unquant Terms and data-and-service notice.