Full Transcript

·YouTLDR

Aktienanalyse in Minuten statt Stunden – mit Claude Code

8:251,006 summary words · ~5 min readEnglishBy The Nexus AITranscribed Jul 14, 2026
Analyze another video with Pro30-day money-back guarantee
Summary

By configuring Claude Desktop with external financial plugins and a structured multi-phase master prompt, retail investors can completely automate professional-grade stock research and generate formatted reports locally in minutes.

This setup shifts the unit economics of security analysis away from expensive enterprise SaaS terminals and manual research teams, enabling individuals to run highly customized, agentic financial pipelines on local hardware.

Section summaries

0:00-1:00

Introduction to Agentic Stock Analysis

watch

The host introduces the concept of using Claude (referred to as 'Cloud Code') to democratize high-level equity research. By leveraging ten specialized financial agents, individual retail investors can execute research tasks that historically required entire institutional research teams. The video promises a step-by-step setup guide for building these custom financial workflows locally using Claude Desktop.

  • Structured AI agents can compress hours of institutional-grade financial analysis into minutes.
  • Ignoring these automated tools represents a massive competitive disadvantage for private investors.

It frames the paradigm shift in research economics from manual labor to local generative agents.

1:00-2:00

Workspace Directory Setup and Spickzettel Generation

optional

The host demonstrates opening Claude Desktop's code editor interface to initialize the project. He commands Claude to generate a dedicated project directory called 'Finances'. Once created, Claude is instructed to draft a 'Spickzettel' (cheat sheet) listing the capabilities of the ten financial plugins, providing a local reference file for the system workspace.

  • Establishing a dedicated local folder structure is the first step to maintaining a clean agentic workspace.
  • Generating a reference document helps anchor the LLM's understanding of its available tools.

This details simple workspace preparation that experienced developers or terminal users can easily replicate without watching.

2:00-3:00

Installing Custom Financial Plugins via MCP

watch

Because Claude Desktop does not natively possess these custom financial tools, they must be manually integrated. The host navigates to the 'Anpassen' (Customize) settings, accesses 'Personal Plugins', and adds a external marketplace repository. This synchronizes the remote GitHub repository with the local desktop environment, opening up a specialized marketplace for financial analysis tools.

  • Enabling live capabilities requires linking Claude to external APIs and custom repositories.
  • Synchronizing repositories allows Claude to run tools locally, bridging the gap between web data and local workspaces.

This shows the exact technical steps needed to configure personal plugins and MCP integrations in Claude Desktop.

3:00-4:00

Selecting Plugins and Crafting the Master Prompt

watch

The host selects key plugins: Earnings Reviewer, Financial Analysis, Equity Research, and Model Builder. He then emphasizes the critical role of a 'Master Prompt' which establishes a strict role-definition, output target, and execution timeline for Claude. This ensures that a single high-level command like 'Analysiere Nvidia' initiates a complete structured pipeline rather than a superficial chat response.

  • A rigorous master prompt acts as the workflow's operating system, standardizing all downstream agentic outputs.
  • Isolating plugins like Earnings Reviewer ensures the agent uses the correct tool for specific steps in the pipeline.

It explains the structural architecture of the master prompt which governs the agent's behavior.

4:00-5:00

Executing the Nvidia Analysis Pipeline

watch

The master prompt is executed with the command to analyze Nvidia. Claude immediately creates a dedicated subdirectory for 'Nvidia' and begins a four-phase analytical pipeline. It extracts current earnings reports, pulls peer performance comparisons, compiles valuation multiples, and begins synthesizing the complete report into the local folder in real time.

  • Phased execution keeps the agent grounded, separating data acquisition from qualitative synthesis.
  • Structuring prompts to automatically generate company-specific folders ensures optimal long-term file organization.

This section illustrates the practical tool execution and live directory generation of the agent.

5:00-7:00

Deconstructing the Markdown Report and Key Risks

watch

The host opens the generated Markdown file ('Earning Summary') in the Nvidia folder. It reveals a highly detailed layout: earnings versus consensus expectations, segment analyses, management highlights, valuation tables, catalysts, and an investment thesis. It explicitly highlights critical risk vectors—such as US export controls impacting Nvidia's access to the Chinese market—and lists all raw data sources for verification.

  • The generated report integrates complex geopolitical risk factors directly alongside quantitative models.
  • Mandatory source listing at the end of the report acts as a guardrail against LLM hallucinations and compliance issues.

It showcases the quality, layout, and qualitative risk depth of the output generated by the automated workflow.

7:00-8:00

Word Document Export and Workspace Flexibility

optional

The host demonstrates that the agent has simultaneously compiled the entire analysis into a clean Word document (.docx) complete with tables and structured layouts. He explains that while this specific template was designed to keep outputs to a highly dense three pages, the master prompt can be adapted to compile 30-page deep-dives depending on the user's requirements.

  • Multi-format output (Markdown for database search, Word for executive viewing) increases the workflow's utility.
  • The length and density of the agentic reports are fully customizable via prompt engineering constraints.

It shows formatting variations and output details but introduces no new setup procedures.

8:00-8:00

Outro and Template Distribution

skip

The speaker concludes the video by summarizing the time-saving benefits of this automated agent setup. He invites viewers to email him for a free copy of the precise system prompts used for the file-structure orchestration and roles. He closes with a standard call to action for likes and channel subscriptions.

  • Replicating this setup successfully relies on having a well-defined folder orchestration system prompt.

Standard video wrap-up with promotional calls to action.

Key points

  • Local Agentic Workflows via MCP Plugins — Integrating GitHub repositories as custom plugins directly into Claude Desktop extends the LLM's architecture with specialized tools like Earnings Reviewers and Financial Analysis engines.
  • File-System Structured Artifact Output — Instead of generating fleeting chat outputs, the agent's master prompt enforces a strict local folder hierarchy, exporting reports in both Markdown for personal knowledge bases and formatted Word documents for presentation.
  • Geopolitical and Qualitative Risk Integration — Beyond pure quantitative multiples, the agent automatically synthesizes qualitative factors such as US export controls on Chinese markets and catalogs them alongside traditional catalysts and segment analyses.
Cloud Code macht Aktienanalysen unfair. Besonders Privatanleger jubeln aktuell, denn jetzt kann Claud Dinge, für die früher ganze Analysten und Research Teams gebraucht wurden. Host
Wichtig bei künstlicher Intelligenz oder bei Sprachmodellen allgemein ist immer, dass man weiß, woher die Ergebnisse dann letztendlich stammen... Host

AI-generated from the transcript. May contain errors.

0:00

Hello and welcome to our video today. Cloud Code makes stock market analysis unfair. Especially private investors are currently cheering, because now cloud can do things that used to be used by entire analysts and research teams. Specifically, it is about 10 financial agents with whom every cloud user suddenly gets their own financial expert.

0:21

a few short tips or summaries, but from the right research and financial analysis with numbers, assessments, comparisons, quarterly figures and much more. What used to take expertise and hours of research, Cloud now makes it understandable and usable for everyone in the shortest time. Ignoring this advantage leaves you with a huge chance and so that you don't miss the whole thing, I'll show you step by step how you can set up your own financial agent with Cloud Code.

0:49

We start the desktop version and you can find the link to it in the video description below. Then we first make sure that we are in the right folder, namely in the code folder. And there we then indicate to Claude that he should first create a project folder for us. In this case, we now create the folder "finances" and enter the whole thing. And we see Claude starts the session and he already creates the project folder that we need.

1:19

So, and since we have already created the folder and the tool has an incredible amount of financial capabilities, I'll tell you now, we should create a kind of top note. So, I've already defined the whole thing. Here the text.

1:34

and will now send the text and see that Claude is already working or will think about it right away and will then execute my command. I have to click here again on "take over changes"

1:55

And now the command has already been executed and worked. This file will later serve as a valuable reference in our project and contains everything you need to know about the new finance plugins. So that we can actually use the new capabilities, we need to know how to install them first. We have now received the feedback that no finance plugins have been installed yet.

2:18

and have two options here that have been suggested I prefer option 2 first install plugins then write the cheat sheet and we can already see that no finance plugins are listed and we have to do everything manually here so now we have to install the missing plugin in english it will be called customized

2:44

And in the case of "yes" it is now "adjust" I have set it to German and there we go to the little plus personal plugins and go to "create plugins" and add "marketplace" when opening the tab. And now we have the option to add a repository. I will also link it in the video description, which repository is important. And then we click "copy and paste"

3:13

and then synchronize the repository with cloud desktop. When the synchronization was successful, a new window opens with all the financial plugins that we can install. For us, earnings reviewer, financial analysis, equity research are now interesting.

3:37

and model images. Now I give Claude my master prompt. I like to make the complete prompt available to you for free. I simply write an email for this. You can find the email address below in the video description. And the prompt is so important because it gives Claude a solid process for action analysis. In the end, a single command such as "Analyze NVIDIA" is enough. Claude automatically creates a complete research folder based on it with sufficient scope.

4:06

The master prompt, the role and the goal, was the first step. And the second step was to determine the file and folder structure for the analysis and artifacts. And I did that deliberately because I would like to have every analysis to be determined by the company name in the folder, that a folder is created for it and then everything appears in the folder that is analyzed.

4:32

In the next step I have now simply set up, analyze Nvidia and then we can already see that the folder has been created and the analysis is carried out. He started with phase 1, i.e. current Nvidia earnings are searched at this point in time and now he is still getting various other data from the Internet, namely the peer data and the evaluation multiples and now

5:00

In the next step, he had searched all the data for phase 42 and now he creates a complete report

5:08

and that should also go relatively quickly. So, when I jump into the folder now, then under Finances I see our plugin here, or rather Cheat Sheet, and then I can already see the NVIDIA folder that was created here. And when I go in there, I can already see the first results that were stored. In this case, I click on Earnings Summary and there I see a brief overview of all the important

5:33

parameters and data of the company. I first look at the short term, core results versus expectations. Then I see the financial profile, development of the last quarter or the last quarter. In this case, from an operational point of view, with a complete master's degree in business administration, I can say the numbers are very, very good and very healthy. The segment analysis

5:58

Then we see management highlights and qualitative statements here. Then we also see the evaluation analysis here, the peer comparison. We see the quick assessment simplified down here and the investment thesis, important risks also ranked. The strongest concerns Nvidia US export controls. Everything about this

6:24

in relation to the Chinese market, which has a very high impact for the company. We are the catalysts and in the end we still get a conclusion and we are still the whole sources. Important for artificial intelligence or for language models in general is always that you know where the

6:43

where the results finally come from, because some AI models also tend to show socially disadvantaged behavior or to hallucinate. We'll take a quick look at the summary again.

6:57

This was also stored as a clear Word document in our folder. We see the date very nicely here in the headline. We see here Nvidia Corporation earnings update and have the parameters mentioned here again in a clearer Word document and also in the form of numbers and tables. And see here it is already very, very well listed and executed.

7:23

and can be very satisfied with the first results. We can already see that everything has been executed. The NVIDIA analysis was carried out completely and is ready in the folder. And we see here again, we have two files, once a Markdown file and once a Word file. I have now deliberately kept it slim because I didn't want to have 30 pages, but only the most important ones.

7:49

parameters summarized on three pages depending on the prompt you can of course make it more specific and detailed that's it for today's video I hope you liked the video if so, please leave a like and a subscription we would be very happy

8:04

And I hope you now know how to use Cloud to carry out financial analysis professionally. And feel free to write us an email, then I'll send you a role and goal master point and also like the file and structure prompt. And I hope you enjoyed the video. See you soon, your Mike.

Continue with YouTLDR

Analyze another video with Pro

Process a new video, search every timestamp, compare sources, and keep the result in your library.

Get Pro — $12/month30-day money-back guarantee

More transcripts

Explore other videos transcribed with YouTLDR.