ModelBest Launches Professional Document Agent Workbench Lantay

Ask AI · How does ModelBest's local data design ensure information security for high-rigor industries?

AI Application Barometer (Official Account: ZhidxcomAI)

Compiled by Bi Weihao

Edited by Moying

Zhidx reported on April 16 that the night before last, ModelBest released Lantay, a Cursor-like professional document agent workbench. It is described as a document plug-in designed for document workers that understands the industry, is secure, and is handy, targeting professionals in high-rigor industries such as lawyers, professional authors, and investment heads.

The official claim is that this is a truly industry-savvy and handy document agent workbench: the all-in-one version ensures sensitive information does not need to go to the cloud; a built-in vertical domain knowledge base guarantees professionalism; multiple workspaces are integrated to solve problems in one workbench; and multiple Agents assist in task execution, making it easy to command agents to work.

However, this is information obtainable from official promotion. After the platform launched its public beta, as a heavy document user, I downloaded and experienced it immediately.

1. Real Experience: A Day Where a Local Folder Becomes a "Smart Workspace"

After opening Lantay, you can choose any folder on your computer to create a workspace. This workspace establishes a link with the local folder, and operations on both sides are synchronized in real time.

For example, if you modify the folder in the Lantay workspace, the local folder will also update synchronously; similarly, if you delete a file in the local folder, the workspace will also delete it synchronously.

I chose a folder summarizing recent articles to experience Lantay's basic functions.

The official promotion mentions this is a Cursor-like platform, adopting the Vibedocing intelligent document interaction paradigm. Vibe is your intent, and Docing is the AI executing it.

Just as developers use AI programming tools for Vibe Coding, in Lantay, document workers can collaborate with AI in real time on the same interface.

This so-called Vibedocing can be felt roughly after entering the workspace creation. Presented before you are three sections: the navigation area on the far left, the workspace in the middle, and the dialogue area on the right.

These three sections cover three daily usage scenarios for document workers: selecting files, editing documents, and AI assistant assistance.

Usually, these three areas are in different locations, requiring switching operations each time. Once the number of related files increases, searching and switching operations become very time-consuming.

Lantay places the three areas on one page, allowing all operations to be completed on this page.

During the trial of basic editing functions, I found that Lantay actually does not support editing operations within the software and will guide you to modify externally. Moreover, after completing editing within the software, the document did not save the changes.

After failing to experience the basic editing functions, the next focus was on the combination of the built-in browser and AI assistant.

First, I asked the AI assistant on the right to format the articles in the Word document, inputting the following prompt:

Optimize the format of the articles in this document, use different font sizes and colors for main and subheadings and bold them, set line spacing to 1.5 times, and delete extra spaces.

Lantay first confirmed my specific requirements, then began calling the corresponding tools to operate, but when it finally announced completion, I did not see the modified document.

Subsequently, I asked it to create a new Word document. Lantay worked for a long time and gave me a Markdown file meeting the requirements, but the Word generation still failed.

▲ Left is the Markdown file, right indicates Word generation failed

The built-in browser is opened by clicking the three points in the upper left corner. The plus sign in the upper right corner can add URLs to web references, and the message symbol on the left can send the corresponding website to the dialogue area.

I searched for materials related to OpenClaw in the browser, then asked it to summarize and generate an analysis report. It can be seen that in terms of reading and summarizing web pages and documents, Lantay performed quite well, information capture was relatively accurate, and the output report content was detailed.

❕ This image appears to have used AI generation technology, please discern carefully

▲ Middle is the generated report, totaling 10 pages

However, there are still some small issues: for example, after opening the built-in browser, Google search requires manually entering the URL to jump, and the browser does not provide translation functions.

There are also two things to note: one is that if the local folder is renamed, the original link will become invalid, and you need to reselect the file to establish a connection;

The other thing is that each workspace has a capacity limit, accommodating at most 200 files, with a single file limit of 100MB, and audio files can reach 500MB. Although there are limits, it is sufficient for most scenarios.

In addition, Lantay supports dual interaction of natural language and voice. It can generate meeting minutes from meeting recordings and organize them into documents. You can record directly during meetings, and Lantay will help extract viewpoints and organize them into documents.

2. Multiple Agents Available, Powerful Plug-ins for Legal, Finance, and Literature Directions

Lantay also has multiple built-in Agents to assist in task execution. You can call the required Agent by entering "#" in the dialogue box.

▲ There are 12 Agents in total, covering legal, finance, and literature fields

The official promotion states the product's applicable fields are law, literature, and finance. So, I selected three tasks existing in reality within these fields to see the specific effects.

First, I uploaded a legal case file, which gathered civil cases involving various legal provisions. I hoped it could help me organize the intellectual property parts of these cases, generating intellectual property case experience summaries and reflections as well as litigation strategies.

During the process, I enabled the Agent named "Lawyer Assistant - New Case Litigation Strategy Analysis". It generated a very detailed experience summary and reflection for me, and also customized litigation strategies.

In literature, I uploaded a file containing 115 academic papers related to "Records of the Grand Historian (Shiji)" and submitted two tasks to it:

Enable "Paper Analysis Assistant" to analyze papers related to Shiji and provide graduate thesis research directions; enable "Literature Review Writing Assistant" to write a literature review about Sima Qian.

Although both tasks took a long time to complete, it can be seen that Lantay strictly executed the corresponding Agent's operation steps, and the analysis of the papers was quite thorough. The total word count of the literature review reached over 10,000.

In writing analysis reports, although the report was relatively complete overall and output valuable viewpoints, the number of analyzed papers was too small and did not cover all Shiji-related papers, so the results have limitations.

In terms of literature review, Lantay wrote quite well overall. The review content was very rich, including nearly 10,000 words on multiple aspects such as Sima Qian's historical view, narrative art, and research on literary groups. The references were all papers already in the file, and there was no fabrication of papers.

However, this function still has some issues. For example, the generated results are not in one document but split into several documents. The first document only contains the main title and main paragraphs, while sub-arguments are placed on separate files, requiring subsequent self-organization.

▲ Sub-argument page

In finance, I uploaded some personal invoices and enabled "Lawyer Assistant - Invoice Organization" to organize my recent travel invoices into one document. Lantay not only summarized the invoice table but also categorized them by travel time, cost statistics, invoice header, and traveler information.

The issue to point out is that Lantay's task processing speed is relatively slow. This varies depending on the task. The speed is very slow when processing bulk papers and legal cases, while the process of organizing articles and invoices is slightly faster.

At the same time, task processing speed using Agents is slower than without calling them. Of course, the output quality also differs. However, although a single task is relatively slow, it can execute tasks in multiple workspaces simultaneously, and closing the workspace will not affect task progress.

3. Custom Agent Canvas, Build What Process You Need Yourself

Lantay also provides a custom Agent function. Users can create their own Agent and edit it in the canvas. For example, I created an article polishing Agent, hoping to achieve the level of modifying articles into news style.

First, click settings in the upper right corner to enter the Agent page, then create a new Agent to open the canvas.

After creating the Agent, you can see a blank canvas with your newly established Agent. Click add step to customize the process for your Agent.

During customization, you can further align the Agent's output results with expectations by adding templates. I added a People's Daily news release as a text template here.

After creating the Agent, it can be used in the dialogue area. I used AI to generate a text based on a news event, then asked Lantay to polish it using my newly created Agent. It can be seen that Lantay executed the task according to my predetermined steps. Although the generated news still has some AI flavor, it is generally readable.

4. Professional, Convenient, and Secure, Three Major Advantages Make Document Workers Love It

According to the article released by the official, Lantay focused on three things from the beginning of design: professionalism, convenience, and security.

In terms of professionalism, Lantay has a built-in vertical professional knowledge base. For example, when a lawyer uses it to process files, Lantay will call the professional knowledge base for retrieval. The retrieval process and sources are clearly visible throughout, traceable and verifiable, ensuring professionalism and precision.

▲ Every execution process is transparent and visible

Lantay also provides an "@File" function. You only need to @ the specified file in the dialogue area to precisely link the specific file, allowing AI to analyze based only on the original text, reducing situations where the answer does not match the question.

In terms of security, Lantay's all-in-one version can ensure sensitive data does not go to the cloud according to customer needs.

However, it should be noted that since it is still in the testing phase, some files will be uploaded to the cloud when encountering bugs. Engineers will see these files during the bug fixing process. If you are concerned, try not to upload sensitive files.

As for convenience, the previous cases have made this very clear. Lantay gathers various pages that require complex switching into one, can batch processing various files, and supports multi-session and background execution, which is relatively convenient.

Conclusion: One Screen, One Workbench, Vibe Coding for Document Workers

Just like the official product positioning, as a Cursor-like document agent workbench, non-developer document workers can also command various Agents to complete tasks in Lantay like Vibe Coding.

After experiencing it, although there are issues such as inability to edit documents, slow generation speed, inconvenient web search, and limitations on file size and quantity, it is still relatively practical overall.

Especially in batch file processing, Lantay performs very well. The generated results can guarantee a certain level of professionalism. At the same time, the full transparency of its task execution process significantly increases user trust in the output results. If functions and experience can be further improved in the future, it should be a good software.

This page provides an editorial summary based on publicly available information. It is not a republished article. Use the source link below for the original report.