Это русская версия материала. Для полноты языковых маршрутов текст основан на существующей основной версии.
Anthropic's GitHub has an additional warehouse called anthropics/financial-services, which broke 11K stars within a few days of going online, averaging 1300 + stars per day.
This is not a demo project.* * It contains 10 complete Agent workflows for the financial services industry * *, from the production of investment bank Pitch to the end of the month, all open source, Apache-2.0 license.
# What's in the warehouse
According to the description of README, this warehouse provides "reference agents, skills, and data connectors" - reference level agents, skills, and data connectors.Scenarios covered include:- * * Investment Banking * *: Pitch Builder automatically generates roadshow materials
- Equity Research * *: Valuation review and multi-method cross-validation
- Private Equity * *: due diligence memos and transaction analysis
- Wealth Management * *: Customer profile and asset allocation recommendations
Each Agent comes with complete system prompts, skill definitions, and data connector configurations.It is not a semi-finished product that "gives you a prompt to play by yourself", but a reference implementation that can be directly put into production and use.
# Two deployment paths
The same set of code in the repository, two deployment paths:
- Claude Cowork Plugin * : Ready to use, suitable for business users to get started with zero configuration.Excel, PowerPoint, and Word have corresponding add-in. * Claude Managed Agents API * *: Deploy behind your own workflow engine via Anthropic's Managed Agents API.The same system prompt, the same set of skills, you decide where it runs.
This "one set of code two ways to run" design is quite clever.Business users want out-of-the-box, developers want controllability and customization.Anthropic does not let either side choose, but makes both sides work.
# Why It's Worth Watching
The adoption of AI in the financial industry has been slow, with compliance and risk control at its core.You don't dare let an AI model directly advise a client on an investment or automatically generate a trade order.
The design of this warehouse is honest: * * It is clear that all outputs require manual review, no investment advice, no execution of transactions, no bookkeeping, no approval to open an account * *.It does an "analyst work product" - an analyst-level draft for review by professionals.This positioning reduces compliance risk and improves actual availability.Financial practitioners do not need to fully trust the judgment of AI, just use it as an efficiency tool.
# The Meaning of Open Source
Anthropic chose Apache-2.0 licensing over closed-source, indicating that they want this warehouse to be the industry reference standard for financial agents.Other companies can fork, change, and build their own solutions based on it.
11K stars reached within a week, indicating that the market demand is real.The application of AI in the financial industry is not a question of "whether to use it or not", but "how to use it to comply".The warehouse Anthropic gives an actionable answer.
Primary source:
- anthropics/financial-services GitHub repository
- Anthropic Official Claude for Financial Services Announcement