Analyst Kit Claude: Turn Your AI into a Verified Investment Analyst (2024)

Analyst Kit Claude represents a significant leap forward for quantitative researchers and institutional investors relying on artificial intelligence. For years, using large language models for financial analysis has been fraught with peril; agents were described as totally unreliable, prone to hallucinations, and often stopped before completing necessary research tasks. They frequently relied on unreliable data sources that could lead to catastrophic errors in portfolio management. Analyst Kit (YC W23) solves these critical issues by packaging workflows and tested code inside skills that Claude Code / Codex can use to perform detailed analysis. This tool transforms a standard coding agent into a professional investment analyst capable of hedge-fund-grade equity research without requiring manual intervention or verification of every single data point. By providing real-time and verified data alongside pre-packaged verified code, the kit ensures that the agent does not have to perform the same functionality again and again, effectively acting as a self-contained library of financial expertise.

What is Analyst Kit?

Analyst Kit is an installable product designed to offer hedge-fund-grade equity-research skills specifically for AI coding agents. It is not merely a plugin that adds a few commands; it is a comprehensive suite that packages complex workflows and tested code inside skills that Claude Code / Codex can execute directly. The core concept revolves around the idea that an AI agent should not be expected to reinvent the wheel or guess at data sources when performing high-stakes financial analysis. Instead, Analyst Kit provides a collection of self-contained folders containing instructions and runnable scripts that an agent loads on demand.

The product is entirely open source and free, making institutional-grade tools accessible to a broader range of developers and investors. It functions by addressing the specific limitations of current AI models: the tendency to hallucinate, the inability to access real-time verified data, and the propensity to stop mid-task. By integrating these skills, users can install them into Claude Code as a plugin or copy them into any agent runtime using the bundled installer. This setup allows the AI to access U.S. institutional 13F-HR filings, U.S. SEC filings (including 10-K, 10-Q, and 8-K), and Taiwan (TWSE/TPEx) market data with a level of accuracy previously unattainable by automated agents.

Why It Matters: Solving Hallucinations and Reliability

The primary reason Analyst Kit matters is the fundamental unreliability of unassisted AI agents in financial contexts. Standard models like Claude or Codex are often criticized for being totally unreliable when handling sensitive financial data. They hallucinate facts, invent numbers, and often use unreliable data sources that can invalidate an entire research report. In the world of equity research, a hallucinated revenue figure or a misinterpreted 10-K filing can lead to significant financial losses. Analyst Kit actively catches these hallucinations and ensures that all data used in the analysis is verified before it is presented.

This tool changes the landscape by enabling high-quality research on companies and building investment research using Claude, which was otherwise extremely difficult to achieve. It fixes the issue where an agent stops before completing research by providing the necessary logic and data access within its skills. Furthermore, it provides uncommon accounting rules and thorough detailed analysis, ensuring that the output meets professional standards rather than just conversational expectations.

For investors, the implication is a shift from "best guess" AI to "verified fact" AI. The kit ensures data quality by using real-time and verified data, which is crucial for timely decision-making in volatile markets. By storing reusable workflows, the system also improves as time goes by, learning user preferences and optimizing performance. This not only saves time but also reduces token costs by avoiding the need to re-explain basic financial logic or re-fetch common data points. The result is an AI that behaves like a junior analyst who has been trained on verified code rather than a chatbot guessing at answers.

How It Works: Skills, Workflows, and Verified Data

The mechanism behind Analyst Kit is built on a modular architecture where complex tasks are broken down into specific skills. Each skill is a self-contained folder of instructions and runnable scripts that an agent loads on demand. This structure allows for modularity; if a user wants to perform a specific type of analysis, they do not need to rewrite the entire codebase. Instead, the agent selects the appropriate skill from the kit.

The workflow generally follows this process:

1. Skill Selection: The user or the agent identifies the task, such as analyzing a specific stock or building a thematic investment strategy.
2. Skill Loading: The agent loads the relevant skill into its runtime. This could be installing it as a plugin in Claude Code or copying the files into any agent runtime with the bundled installer.
3. Data Verification: The skill accesses the necessary data sources, which are pre-validated. The system ensures that all data used is verified, preventing the use of outdated or fake information.
4. Execution: The agent executes the runnable scripts, which handle the heavy lifting of data retrieval and processing.
5. Output Generation: The final output is generated in the required format, whether it be a normalized ranked CSV, Markdown, a branded PDF (A4 portrait or 16:9), or a self-contained HTML page.

The kit includes a wide array of capabilities, including 13f-analysis, sec-filings, financialmodellingprep, finmind, market-intelligence, company-universe-manager, analyzing-financial-statements, creating-financial-models, charting, reporting, wiki-builder, data-analysis, and more. By utilizing these specific skills, the agent can navigate the complexities of financial modeling without needing to understand every line of code from scratch. The system is designed to be transparent; users can see exactly what data is being used and how the analysis is being constructed.

Use Cases: From Deep Dives to Thematic Investing

Analyst Kit is versatile enough to support a variety of investment strategies, moving far beyond simple price checks. One of the most powerful use cases is the single-stock-deep-dive. In this scenario, an analyst can command the AI to perform a comprehensive review of a company's financial health. The agent will pull the latest 10-K and 10-Q filings, analyze the financial statements using verified accounting rules, and create a financial model to project future performance. The output will be a detailed report that catches hallucinations and ensures the data is real-time.

Another significant application is thematic investing. Investors often want to explore sectors like renewable energy or artificial intelligence. With Analyst Kit, the agent can utilize the thematic-investing workflow to scan the market for trends, analyze the competitive landscape, and identify key players. The market-intelligence skill allows the agent to gather data on multiple companies simultaneously, facilitating a broad overview of a sector.

The kit also supports the creation of an analyst-playbook. This is particularly useful for firms that want to standardize their research process. By storing reusable workflows, the system learns user preferences over time, ensuring that the playbook evolves with the firm's specific needs. For example, if a firm prefers a specific format for their PDF reports or a particular set of metrics for their CSV exports, the system adapts.

Furthermore, the Analyst Kit can be used for company-universe-manager tasks. This involves managing a watchlist of companies, automatically updating their data, and flagging any significant changes in their SEC filings. The system can generate a wiki-builder output, creating a living documentation of the research process and findings. Whether the goal is creating-financial-models for a private equity deal or charting historical performance for a public equity, the skills are pre-packaged and ready to go. The result is a level of detail and accuracy that rivals human analysts, but at a fraction of the time and cost.

Installation and Requirements

Setting up Analyst Kit is designed to be straightforward, accommodating various technical environments. Users can install these skills into Claude Code as a plugin, which provides a seamless integration experience within the IDE. Alternatively, users can copy the skills into any agent runtime with the bundled installer, offering flexibility for those who prefer different environments.

The technical requirements depend on the specific runtime being used. Common examples include Python with the standard library, Python with numpy/pandas, Node, and Bun. These runtimes are necessary to execute the runnable scripts contained within the skills. Users will also need to obtain specific API keys to unlock the full potential of the data sources. Required keys include FMP_API_KEY for financial modeling prep, FINMIND_TOKEN for specific financial data access, and SERPAPI_API_KEY for search engine integration.

While the brief does not name specific competitor products, the approach of Analyst Kit stands out by being open source and free. Unlike proprietary enterprise software that charges high licensing fees, this tool is accessible to individual researchers and smaller firms. The open-source nature also means that the code is transparent, allowing users to audit the logic and ensure that the verification processes are sound. This contrasts with "black box" solutions where the data sourcing and analysis methods are hidden.

When comparing the installation method to traditional manual research, the automation provided by Analyst Kit eliminates the need for researchers to manually download PDFs, scrape websites, or cross-reference data points. The self-contained folders of instructions mean that the setup is a one-time effort that pays dividends over time. The system stores reusable workflows, so once a specific analysis is run, the logic is saved and can be applied to similar companies or sectors with minimal additional effort.

FAQs

Is Analyst Kit suitable for beginners in financial analysis?

Yes, Analyst Kit is designed to make complex financial analysis accessible. By packaging workflows and tested code inside skills, it removes the need for deep programming knowledge. Users can simply install the skills and let the agent handle the heavy lifting. The kit provides real-time and verified data, ensuring that even beginners can produce high-quality research without needing to verify every number manually.

How does it handle data accuracy and hallucinations?

The kit actively catches hallucinations and ensures that all data used in the analysis is verified. It uses real-time data sources and pre-packaged verified code so the agent does not have to perform the same functionality again and again. This means that the output is reliable and based on facts, not guesses. The system is specifically built to address the issues where standard AI models are described as totally unreliable.

Can I use this for international markets outside the US?

The kit currently supports U.S. institutional 13F-HR, U.S. SEC filings (10-K, 10-Q, 8-K), and Taiwan (TWSE/TPEx) market data. If you are interested in other international markets, you may need to check if the specific skills cover those regions or if you need to configure the API keys accordingly. The system is flexible, but the available data sources are defined by the skills included in the open-source package.

Frequently Asked Questions

Can I customize the skills for my specific firm's needs? Yes, because Analyst Kit is open source and free, you have full control over the code. You can modify the self-contained folders of instructions and runnable scripts to fit your specific workflows. The system also learns user preferences over time, adapting to your firm's unique requirements as you use it.

What happens if the AI still makes a mistake? The kit is designed to minimize mistakes by actively catching hallucinations and using verified data. However, like any tool, it is important to review the outputs. The system ensures that the data is verified, but human oversight is still recommended for high-stakes decisions. The kit improves as time goes by, learning from corrections and refining its accuracy.

Is there a cost associated with using Analyst Kit? No, the product is free and open source. There are no licensing fees or subscription costs. You only need to obtain the necessary API keys (such as FMP_API_KEY, FINMIND_TOKEN, SERPAPI_API_KEY) to access the data. This makes it an incredibly cost-effective solution compared to hiring a full-time analyst or purchasing expensive enterprise software.

Sources

• Analyst Kit (YC W23): Turn your Claude / Codex into an investment analyst (Free) — Hacker News
• Analyst Kit (YC W23): Turn your Claude / Codex into an investment analyst (Free) — Hacker News

Recommended AI Tools

Sider AI — All-in-one browser AI sidekick that lets users chat, summarize webpages/videos, translate pages, explain text, research faster, and use multiple AI models in one sidebar. Includes Wisebase knowledge...