Brokerage analysts collectively share "Lobster" cultivation tutorial: How does the phenomenal AI agent OpenClaw stir up the investment research circle?

In the finance research and investment circle, the application of large models is no longer new. The excitement isn’t just about the popular scene of queuing outside Tencent Building in Shenzhen to install “Little Lobster” (OpenClaw - AI intelligent agent), but also about collective actions by securities firms.

Recently, the financial engineering teams of several brokerages—including Founder Securities, GF Securities, Dongwu Securities, Northeast Securities, and others—have simultaneously released special reports, guiding financial practitioners step-by-step on how to deploy and use an open-source AI agent called “OpenClaw.”

These reports no longer just discuss macro algorithms but teach investors how to configure servers, install Skill packages, and even set up “private domain AI assistants” on their computers. From installation guides to practical scenarios in research and investment, they form a comprehensive industry application manual.

Brokerage Quantitative Analysts Share “Lobster Farming” Tutorials

OpenClaw, an AI agent with a lobster icon, exploded on GitHub in early 2026, quickly surpassing 150,000 stars and earning the industry’s nickname as “the most explosive open-source project of 2026.” Unlike ordinary users who are keen on its automation features, brokerage quantitative teams keenly recognize its huge potential in research and investment, leading to deep exploration and public sharing of results.

By early March, eight brokerages—including Founder Securities, GF Securities, CITIC Securities, Dongwu Securities, Northeast Securities, and others—had published related research reports. These cover multi-platform deployment solutions for Windows, Mac, and cloud servers, as well as hands-on tutorials for core research scenarios like financial data access, conditional stock selection, financial statement analysis, and quantitative backtesting.

Founder Securities: “Empowering Financial Research with OpenClaw: 17 Efficient Application Cases Explained”

Guojin Securities: “Building a Personal Research Assistant with OpenClaw (Part 2): Skills Setup and Research Work Cases”

GF Securities: “Multi-Platform Deployment and Research Applications of OpenClaw”

CITIC Securities: “Shrimp Farming Guide: Deployment and Experience of OpenClaw”

Dongwu Securities: “In-Depth Evaluation and Application Guide for OpenClaw”

Zheshang Securities: “Next-Generation Research Infrastructure: From Deployment to Application of OpenClaw”

Huachuang Securities: “Using OpenClaw to Build Your Own Private Domain AI Assistant”

Northeast Securities: “Install These 20 Skill Packages on Your OpenClaw to Boost Research Efficiency by 10 Times”

What Makes OpenClaw Excite the Financial World?

To understand this research circle frenzy, first recognize what OpenClaw is. This open-source AI project, featuring a red lobster icon, took the global tech scene by storm at the start of 2026 with unmatched momentum.

Why are brokerage quant teams so uniformly focused on an open-source project? The answer lies in its fundamentally disruptive logic.

“OpenClaw’s core difference from cloud-based large models is where the ‘brain’ is located and whether a ‘hand’ exists,” Huachuang Securities vividly explains in their report.

Traditional cloud-based large models are like omniscient remote advisors, providing only text-based solutions; whereas OpenClaw runs entirely locally or on private clouds, with system permissions equal to the user. It can directly operate the computer terminal, write code, manage files, and even autonomously learn and install new “Skills” based on natural language commands.

Dongwu Securities even defines it as a new generation AI agent, believing it is evolving from a “question-and-answer tool” to a “practical work assistant capable of execution.” This leap from passive response to active execution perfectly addresses the pain points of tedious, high-frequency, multi-tool collaborative work in finance research.

Moreover, the daily volume of structured and unstructured data generated in global financial markets has reached petabyte levels. For frontline research personnel, the amount of information they need to process daily has skyrocketed from hundreds of items ten years ago to tens of thousands today. The emergence of OpenClaw provides a “productivity lever,” exactly what financial practitioners overwhelmed by “information overload” desperately need.

The “Super Employee” of Research: Freeing Analysts’ Hands

Among these reports, what stands out most are the real research and investment scenarios demonstrated. OpenClaw’s capabilities in these scenarios have gone beyond a simple tool, resembling a tireless “super employee.”

1. Automated Information Processing and Market Monitoring

Financial markets change rapidly; analyzing massive announcements and news is daily work for analysts. Guojin Securities demonstrated how to build a “Daily A-share Announcement Summary and Scheduled Dispatch” Skill with OpenClaw. It can automatically fetch announcements, classify and identify key figures and entities, generate Excel summaries, and produce briefings in one sentence, even scheduling daily pushes to analysts’ phones.

Founder Securities also showcased similar market monitoring and conditional stock selection functions, where AI can directly execute stock picks based on instructions and display complete results.

2. Deep Report Writing and Research Reproduction

Dongwu Securities’ tests show that with a simple command, OpenClaw can autonomously access data, write an analysis report including valuation, high dividend yield, and leading effect, and automatically save it as a Word document in a specified local directory. It can also efficiently read and organize fund managers’ research notes, extracting core insights.

Guojin Securities further uses OpenClaw for “automated research report reproduction.” By inputting a research report, it can analyze logic, fetch data, write code for strategy backtesting, and output standardized reproduction results with net value charts and deviation analysis.

3. Quantitative Strategy Development and Database Connectivity

For quant teams, strategy development is core. Founder Securities demonstrated how to grant OpenClaw access to a proprietary factor library, enabling it to autonomously backtest a complex small-cap value stock strategy and filter the latest holdings. Dongwu Securities shared advanced techniques like encapsulating custom Skills to connect directly to SQL databases, execute queries, and extract structured data such as market prices.

By integrating with instant messaging apps like Feishu, DingTalk, and Telegram, analysts can now simply send a voice message in chat, and OpenClaw on remote servers will quietly execute complex research tasks.

Security and “Illusions” Still Hang Over the Sword

Despite OpenClaw’s impressive productivity, brokerage analysts all included prominent risk warnings at the end of their reports.

Dongwu Securities warned about permission risks: since OpenClaw has “super permissions” over the operating system, improper configuration or unreliable third-party Skill packages could lead to accidental deletion or leakage of important local files.

Founder Securities mentioned AI hallucinations: even if OpenClaw can fetch data and write code automatically, underlying logical errors or fabricated data in large models cannot be completely eliminated. For finance, which demands logical rigor, AI-generated conclusions should only serve as “auxiliary references,” with final approval firmly in human hands.

Additionally, because OpenClaw has system-level permissions, it can read/write files and execute terminal commands. Huachuang Securities issued a warning: “Strongly recommend not installing or using it on your main work computer.” Founder Securities also emphasized that OpenClaw could pose significant threats to local environments and files, and strongly advised deploying it in isolated environments separate from work or personal computers.

In a secure, isolated environment, granting AI limited database access can greatly reduce errors caused by “hallucinations” and maximize its productivity.

Research Work Moving Toward Intelligent Transformation

The collective effort of financial engineering analysts in writing deployment reports for OpenClaw reflects not just a technical curiosity but a broader industry shift toward intelligent transformation.

Guojin Securities pointed out that the significance of OpenClaw lies in helping research personnel upgrade from “ad hoc” to “stable” usage. This means OpenClaw is no longer a tool that forgets everything after each interaction but a long-term digital avatar that can remember analyst preferences and evolve over time.

By standardizing scattered, repetitive research processes into Skills, the research system is evolving toward a structured, reusable, auditable framework. AI automatically searches, downloads, configures, and tests, internalizing these as permanent capabilities. This skill internalization, combined with its Workspace-based “long-term memory,” makes OpenClaw smarter with use.

As Dongwu Securities states, technological upgrades have already achieved leapfrog efficiency improvements. The new generation of intelligent tools represented by OpenClaw is rapidly reshaping the underlying logic and practical paradigms of research work.

Of course, current OpenClaw still faces challenges such as high usage barriers, fast token consumption, and an underdeveloped ecosystem. Meanwhile, all brokerages remind that AI-generated conclusions are only for reference and cannot replace independent judgment, in-depth analysis, and final decision-making by professional research personnel.