Moonshot launches Kimi for finance with 10+ data feeds and 9 skills

Moonshot AI launched a financial industry package for its Kimi models on Thursday, wiring the assistant into more than 10 market data sources, nine finance-specific skills and five compliance controls. The company’s own announcement names Wind, East Money, S&P Global, Cailianshe and Caixin Data among the feeds. It says dozens of financial institutions are already building on Kimi models, Kimi Enterprise and the Kimi Work desktop product.

CNBC reported that investment bank CICC and HongShan, the venture firm formerly known as Sequoia China, are among the users. Crunchbase and the business database Tianyancha are also on the partner list, and Kimi can pull filings straight from the SEC’s EDGAR system as well as IMF, World Bank and FRED data. A CNBC test of the mobile app found that the data on offer varied by subscription tier. Subscriptions start at 49 yuan ($7.31) a month and run up to 699 yuan ($104.23).

PANews, citing Xinhua Finance, says the first batch of deployments includes ICBC, CSC Financial, CICC and E Fund. Moonshot itself says the users span leading brokerages, commercial banks, venture funds and research data providers, which means it’s pitching across the buy side and sell side at once rather than picking a niche.

The mechanism is tool calling. Moonshot says the data feeds plug into research tasks through MCP, so a figure in a draft can be traced back to its original source and checked. On top of the feeds sit nine packaged skills: institutional financial modelling, institutional research reports, institutional slide decks, dynamic financial charts, earnings commentary, a consensus expectations map, portfolio review, a morning holdings report and a Hong Kong IPO lens. The company published four anonymised case studies, and says none of them corresponds to a real company.

Task (Moonshot’s own case study)	Before	With Kimi
Financial model for a commercial space company	5 to 7 person-days	0.5 to 1 person-day
Deep research report on quantum technology	10 to 20 days	About 2 days
Bank credit investigation report	Baseline	Over 5x faster per borrower
21-page pet insurance investment deck	7 to 10 person-days	About 2 person-days

Those are vendor figures, and Moonshot doesn’t name the institutions behind them. The one named pilot is with CSC Financial, where the two companies built what they call a risk assessment gateway. It enforces five controls: data classification, personal information filtering, per-call authorisation of data sources and tools, source-labelled output with human review, and an audit trail linking each task to its data and tool calls. CSC used it to draft more than 60 interim bond trustee reports covering over 50 issuers and more than 100 outstanding bonds.

Moonshot says the integration was planned to take two months and took three working days, and that drafting time per report fell from about 30 minutes to 10, a 67% cut in manual work. Business staff kept the review and sign-off role. Even so, nobody outside the two companies has checked those numbers.

AI companies that understand real financial workflows and can deliver reliability and data security will be particularly well positioned to contribute to this transformation.

Samuel Fischer, Beijing branch manager at Deutsche Bank, in a Moonshot promotional video, via CNBC

Deutsche Bank’s presence in the launch video doesn’t make it a client. CNBC said it wasn’t immediately clear whether the bank uses Kimi, and the bank didn’t respond to a request for comment. The catch is that the data security Fischer praises is exactly where Moonshot took a hit last week.

Anthropic’s September threat report says Moonshot silently forwarded customer requests to Claude, showed users Claude’s answers as if they came from Kimi, and saved the exchanges to train its own models. In one ten-day period it relayed almost 300,000 customer requests through a network of 5,380 fraudulent accounts, and Anthropic counted over 23 million exchanges between May and July. Some of the rerouted queries carried sensitive customer data, including live credentials from major Chinese companies. Moonshot’s financial services announcement doesn’t mention the report.

That matters because the whole pitch to banks rests on knowing where your data goes. The gateway’s per-call authorisation and audit trail are the answer Moonshot is offering, but they govern what a bank’s data may reach. They don’t speak to the question Anthropic raised, which is where a request goes once Moonshot has it.

The money explains the push. TechCrunch, relaying a Bloomberg report, says Moonshot is targeting $2 billion in annualised revenue by the end of the year, double its reported run rate for August. The Kimi K3 model that drives that growth shipped in July, and OpenRouter data cited by TechCrunch shows as many as 300 billion tokens a day generated by K3 models on that platform. Open weights keep Moonshot’s margins thinner than closed rivals, though, and its projected revenue is still a fraction of Anthropic’s $65 billion run rate.

Finance is where a Chinese lab can charge for workflow rather than tokens, after a year in which its open-weight peers mostly competed on price, as Together AI’s GLM-5.3 comparison showed. Moonshot has reportedly filed confidentially for a Hong Kong IPO, and CNBC noted it doesn’t comment on market speculation. What to watch is whether a named Western bank signs on, and whether the distillation findings surface in that filing.