The financial data MCP landscape is fragmented.

What this benchmark measures

We inspected public, independently retrievable surfaces on August 1, 2026. A linked or “Observed” entry means that a relevant public page or listing was found; it does not prove product quality, data accuracy, endorsement, traffic, or inclusion in a model's training data. “Not observed” means only that this audit did not find the surface, not that it is universally absent.

The benchmark focuses on distribution. Registry entries, public repositories, detailed HTML documentation, and regional marketplaces create more paths for people and agents to find a product.

Four findings

1. The Official MCP Registry is a multiplier, not a complete market map. It gives aggregators and clients a canonical metadata source, but many established products remain discoverable primarily through their own docs, GitHub, or community directories.
2. A hosted remote server does not need an npm or PyPI wrapper. The Registry supports a public Streamable HTTP endpoint directly. Public examples still matter because GitHub creates an inspectable setup and citation surface.
3. China requires separate distribution. Chinese documentation, Baidu's AI MCP plaza, ModelScope, Tencent surfaces, and mainland performance are distinct from a Western registry listing.
4. Machine metadata cannot replace substantive HTML. Tool schemas and llms.txt help agents connect, while dataset pages, methods, limitations, and original research give search and answer engines material they can index and cite.

Where Unquant fits

At the audit date, Unquant already operated a public remote MCP with 28 described tools, including 17 live data tools, explicit safety annotations, and a broader research catalog. It allowed major search, answer-engine, and model-development crawlers. Its visibility gap was distribution: only five HTML URLs in the sitemap, no Official Registry listing, no public examples repository, and no native Chinese content.

This release adds public tool and dataset catalogs, detailed dataset pages, methods, security information, and English and Chinese research guides. After the audit, Unquant published a public examples repository with a fixed commit and a passing validation workflow. The Official Registry release is separate and needs its own receipt. A listing does not prove crawling or indexing. Each event needs separate verification.