Helper

What is a Helper?

Helper is a cross-platform desktop AI assistant designed for users who need to quickly utilize large language models in tasks such as writing, programming, marketing, and everyday office work. It supports macOS, Windows, and Linux, and can be invoked from the current application using global shortcuts.

It is not merely a wrapper for a single chat page; rather, it integrates conversations, prompts, AI commands, custom assistants, and text responses into a desktop workflow. Users can choose the Cloud option, which utilizes the models provided by the platform, or they can use the Local option to configure their own model service keys.

Current product status and version information

As of this verification, the latest desktop version displayed on the official download page is v1.3.2, with an update date of January 24, 2024. This version includes multiple engines and over thirty different model options, Markdown formatting, code highlighting, as well as the free Gemini Pro model available at that time.

Since the last public update was made a long time ago, the model names listed on the older versions of the pages should not be considered to represent the current latest models. Before downloading and installing, it is necessary to check again the file version, system compatibility, and any security warnings.

Core functions

Global shortcut to activate

Users do not need to switch between browser tabs frequently; they can open the Helper using shortcut keys in applications such as documents, emails, and code editors. On macOS, the default shortcut is Option plus X, while on Windows and Linux it is Alt plus X.

Three types of output: chatting, typing, and pasting

AI Chat is suitable for continuous questioning; the typewriter mode writes content as it is generated within the current application, while the paste mode waits until generation is complete before inserting the content all at once. These three methods offer different trade-offs between speed, format preservation, and continuity of interaction.

Prompt words, commands, and custom assistants

By entering a slash, a menu of prompt options is displayed; users can also reuse the built-in prompt library, AI commands, and preset assistants. The official website claims to offer more than 50 AI commands as well as over 50 AI assistants.

Multiple models and built-in keys

Helper can connect to model services such as OpenAI, Gemini, Azure, and OpenRouter, and in local setups it allows users to configure their own API keys. The availability of the models, the maximum length of the context, and the costs depend on the service provider chosen.

Markdown and code display

The dialog window supports Markdown formatting and code highlighting, making it suitable for viewing structured responses and code snippets. When the content is exported to an external editor, the final format will still be determined by the capabilities of that software.

List of main functions

• Activate the AI using global shortcut keys in different desktop applications.
• Consecutive conversations and contextual follow-up questions centered around the same task.
• Type or paste the results back into the current window.
• Repeat workflows using built-in prompts, AI commands, and custom assistants.
• Connects to model platforms such as OpenAI, Gemini, Azure, and OpenRouter.
• In the Local option, you use your own API Key and are responsible for covering the costs associated with the models.
• Displays Markdown content, code blocks, and syntax highlighting.
• Some network or enterprise outbound requirements are met through proxy configuration.

What use cases are suitable?

• Writer: Rewrites paragraphs, organizes outlines, condenses long texts, and generates titles.
• Developer: Explain code, generate comments, troubleshoot errors, and outline testing approaches.
• Marketers: Create ad copy, social media content, and campaign ideas.
• Office users: drafting emails, summarizing meeting minutes, and adjusting the tone of expressions.
• Researchers: Ask questions based on existing text and generate structured summaries.
• Multi-model users: Switch between different service providers and models from a single desktop interface.
• API users: wish to retain the ability to manage billing through their own accounts and to choose the models to use.

What situations are not very suitable

• Teams that require continuous access to the latest models and frequent updates from clients.
• Sensitive businesses that involve sending prompt content to third-party model services cannot be accepted.
• Organizations that require official open-source code, auditable builds, or full private deployment.
• Users who wish to make a purchase only once and do not want to bear the costs associated with subsequent API calls.
• Users of devices that rely on the native Windows ARM installation package.
• People who need to use mobile office services on smartphones and tablets with the same client application.

Helper Usage Tutorial

Download and install

1. Go to the official download page and select the installation package that is compatible with your processor and operating system.
2. Check the version number, the source of the file, and the system security warnings before proceeding with the installation.
3. After the first launch, choose between Cloud or Local mode based on actual needs.
4. Allow only the necessary shortcut keys or accessibility permissions; do not grant any unnecessary permissions.
5. Test whether evocation, generation, and filling work properly in non-sensitive text.

Call in any application

1. Open the document, email, or editor containing the text that needs to be processed.
2. Select the existing text, or place the cursor where you want to enter text.
3. Use the default global shortcut to invoke the Helper.
4. Enter the task requirements, or type a slash to select predefined prompts.
5. Choose the output method—chat, typewriter, or paste—based on the task.
6. Read the results again and verify the facts, figures, citations, and code before saving.

Configure your own API Key

1. First, create a key at a supported model service provider and check the account balance.
2. In the Helper settings, select the appropriate AI engine and interface configuration.
3. Paste the key and set the model or trusted proxy as needed.
4. Use short prompts that contain no sensitive information to complete the connection test.
5. Regularly check the service provider’s usage, and rotate or revoke keys as necessary.

Comparison of the three output modes

How to choose between cloud and local solutions?

Cloud package prices

The following are the reference prices and limits displayed on the official website during verification; the currency is the US dollar. The actual settlement amount, taxes, scope of the models, and rules for conversion are as specified on the purchase page.

The page indicates that one invocation of GPT-4 is equivalent to 30 basic queries; therefore, the stated quota cannot be directly equated with the number of times advanced models can be used. The annual price is the monthly average shown on the page, and it depends on the payment page whether the fee is charged in a lump sum for the whole year.

Price of Local package

The Local solution relies on one-time client authorization and includes free updates for one year; the costs associated with the user’s own API calls are not included in the authorization fee. The mention of permanent use on the page does not mean that new versions will be available indefinitely.

The official purchase instructions state that credit cards can be used, and options for WeChat and Alipay are available for Chinese users. Refunds can be requested within seven days for unused purchases, but whether a refund will be granted and the specific procedures depend on the information provided on the payment page and confirmation from customer service.

Platform and system requirements supported

API proxy and network configuration

By default, the Local mode allows connection to the official model interface; in restricted networks or corporate network environments, a proxy can also be configured. The provider also offers a proxy endpoint, claiming that it will forward data without storing it.

In theory, proxy operators can see the content of requests as well as the authentication credentials; therefore, proxies of unknown origin should not be used in sensitive scenarios. Creating one’s own reverse proxy gives more control, but it also entails the responsibility of managing servers, logs, certificates, and keys.

Privacy and data security

The website’s privacy policy states that registration data, support messages, IP addresses, browser information, and access logs may be processed, and that this data is used for maintenance, analysis, communication, marketing, and fraud prevention. The policy describes the activities carried out on the website, but it does not provide a complete explanation of all the ways in which prompts and API keys are handled in desktop applications.

• Do not submit passwords, identification documents, or confidential business information to public models or unknown agents.
• Give priority to using API keys with restricted permissions, which can be revoked and for which usage limits are set.
• Distinguish between website account data, Helper cloud data, and the data received by model service providers.
• Before using the team, it is necessary to confirm the data retention policies, training procedures, cross-border data transfer rules, and log coverage.
• Before deleting the client, check whether a valid key still exists in the model provider’s backend.

Tips for installing security and antivirus software

The official help documentation states that Windows security software may generate false alarms or block certain actions due to the client’s invocation of system-level interfaces. A false alarm does not mean that a file is necessarily safe; therefore, one should not turn off the antivirus software or firewall just in order to install a particular program.

1. Download the installation package only from the official download page, and verify the version as well as the digital signature.
2. Use system security tools and reliable scanning services to check the files.
3. If it is still blocked, first verify the document hash and obtain an explanation for the false positive with the authorities.
4. Once credibility is confirmed, only minimum scope exceptions are set for specific procedures.
5. Stop using it immediately if abnormal networking, privilege escalation, or automatic startup behaviors occur.

Product advantages

• One client supports macOS, Windows, and Linux desktop platforms.
• Global shortcut keys reduce the need to switch between business software and chat pages.
• Chat, stream typing, and full paste options cater to various output needs.
• It offers two options for use: Cloud and a built-in API key.
• Built-in prompts, AI commands, and assistants reduce the cost of repetitive configuration.
• It supports multiple model channels, making it easy for users with existing API accounts to make a choice.
• Markdown and code highlighting improve the reading experience of technical content.

Usage restrictions and precautions

• The public update records go up to 2024, and it is necessary to keep an eye on the level of maintenance activity.
• The local licensing fee does not include the costs associated with using third-party models such as OpenAI.
• The number of Cloud queries is calculated based on the model; more advanced models may consume resources more quickly.
• The Free page shows both monthly and daily quota figures; the amount displayed on the account shall prevail.
• Typewriter mode can ruin indentation or formatting in code editors.
• Agents add an additional layer of entities that come into contact with the data, which makes them unsuitable for sensitive tasks that have not been evaluated.
• The website’s privacy policy does not cover the process of handling desktop prompts in its entirety.
• The Windows version may trigger interference from security software; therefore, it is necessary to verify the files carefully.
• AI-generated content may contain factual errors, incorrect codes, and inappropriate expressions.

GitHub and the open-source status

As of the time of verification, no official source code repository that could be accessed through Helper’s official website and whose ownership could be confirmed was found. The projects with similar names that appear in real-time searches are mostly unrelated tools or third-party clients, and therefore cannot serve as evidence of Helper’s open-source nature.

Its end-user license terms prohibit modification, decompilation, redistribution, and resale; therefore, the desktop client should be classified as closed-source commercial software. The term “local” in the Local solution does not imply that the model or client is open source.

Basic information

Recommendation score

4.0 / 5. The ability to invoke Helper globally, the various ways in which data can be filled in, and the built-in API keys make it suitable for frequent writing and programming tasks on desktops; however, the lack of regular updates, limited information regarding privacy issues, and additional costs associated with APIs pose challenges when considering this option over the long term.

Frequently Asked Questions

Is Helper free?

The Cloud version offers a limited number of free queries, while the Local client requires a one-time purchase of a license. Local users are also responsible for covering the API costs associated with the model service providers.

Which systems does Helper support?

The official version provides installation packages for macOS Intel, macOS Apple chip, Windows x64, Linux x64, and Linux ARM. No official mobile version is available.

What is the difference between Cloud and Local?

Cloud provides access to models and query quotas through the platform, making it suitable for immediate use; Local requires users to configure their own API keys, with separate charges applied for client authorization and model invocation.

Can it be used in any application?

Most desktop applications that allow text input can be invoked using global shortcuts, but issues related to data restoration, formatting, and shortcut conflicts vary from one software to another. It is recommended to test them first in plain text.

Does Helper save the chat content?

The official agent claims that it forwards data without storing it, but the website’s privacy policy does not cover all data flows related to the desktop application in detail. The specifics also depend on the cloud services, the agent, and the model provider chosen.

What should I do if antivirus software gives an alert?

Do not turn off security software directly. First, verify the source of the download, its signature, and the results of the scans; if there are false positives, set minimum exceptions only for programs that have been verified as legitimate.

Is Helper open-source software?

No. No verifiable official source code repository was found, and the licensing terms prohibit decompilation, modification, and redistribution.

Do I still need to pay after making a one-time purchase?

The Local license includes one year of free updates, but the model APIs are charged according to the provider’s rules. The policies and prices for subsequent upgrades after one year will be as specified on the relevant page at that time.