Source: TMTpost Media

In April, not long after Baidu released Wen Xin Yi Yan, many people were still lamenting how happy the pictures generated by Wen Xin Yi Yan were. Even more people were going crazy for various trainings such as ChatGPT and Midjourney. Meta founder and CEO Zuckerberg is thinking about the opportunity to introduce AI Agents to billions of people around the world “in a useful and meaningful way.”

In May, when OpenAI completed a new round of $300 million in financing, founder Sam Altman privately told some developers that he hoped to build ChatGPT into a personal work assistant. Sources familiar with the matter revealed that OpenAI has been paying attention to how to use chatbots to create autonomous AI Agents, related functions are likely to be deployed in the ChatGPT assistant.

At an all-staff meeting in June, Zuckerberg announced a series of technologies in various stages of development, one of which would bring AI Agents with different personalities and abilities to provide assistance or entertainment to users.

Just in July, Meta released the AI Agent project MetaGPT, which is an automatic agent framework focusing on software development based on GPT-4.

In China, although AutoGPT has become popular as early as April in foreign countries, due to the lack of understanding of most people about the AI Agent behind it, the initial response was not too enthusiastic.

It was not until the blog post about AI Agent by Lilian Weng, the head of OpenAI’s applied artificial intelligence research, in early July that the AI circle exploded, that the media, academic and research circles, and investment fields really began to discuss AI Agent enthusiastically.

As a result, the country has really started an upsurge in exploring and researching AI Agents, and some manufacturers have begun to reconstruct product architecture and business models based on the AI Agent model.

As the principles, models, and construction methods of AI Agent become more and more clear, many entrepreneurs who are trapped in technology, models, ecology, and even policies are seeing a bright future.

AI Agent not only allows everyone to see the direction of large language model (LLM, Large language Model), it also allows more entrepreneurs to further ignite the hope of LLM entrepreneurship, and also allows the majority of enterprises to see the future trend of efficient application of LLM.

Regarding AI Agent entrepreneurship, OpenAI co-founder Andrej Karpathy believes that ordinary people, entrepreneurs and geeks have more advantages than OpenAI in building Agents, and everyone is in a state of equal competition.

On the side of large companies, facing the possibility that large technology companies and startups will seize the opportunity of Agent, Bill Gates also said that he would be disappointed if Microsoft did not intervene.

With the strong promotion of technology giants, the rapid embrace of entrepreneurs, and the active introduction of large enterprises, AI Agent has become completely popular. And unlike the previous situation where LLM lacked implementation, this time AI Agent is no longer just a paper idea. Many companies have already launched Agent projects and related products.

Industry insiders revealed that at least 100+ projects are working on commercializing AI agents, and nearly 100,000 developers are building autonomous agents. Among these AI Agents, there are foreign Agent projects mainly based on GPT and open source Agent framework, as well as domestic Agent products based on domestic large models (large models in self-research fields) + open source architecture.

Having said all that, which companies have launched Agent products? What is the current form of AI Agent products? This article counts 60 AI Agents around the world to give everyone a better understanding of AI agents.

**PS: **Because there are many Agent projects reviewed in this article, the number of words has reached 1W+. It is recommended that you collect it first and then read it.

Start with AI Agent

Although LLM has enough intelligence, if you want it to give accurate answers, it needs to be input accurately enough. If a master and an ordinary person use the same large model to ask questions, the answers they get will be very different: the former can use a variety of techniques to get the desired results, while the latter can only look to LLM and sigh.

If you want to use LLM well, you must first learn to use it. This demand has spawned a large training market. The prompt project, while increasing the difficulty of using LLM, also reduces the user experience. LLM, which should have fully demonstrated the advantages of natural language, has become not so friendly to ordinary users because of its complexity.

In this way, the prompt project has become a big mountain between ordinary people and large models.

How to better solve this problem? The answer is AI Agent (called AI agent in China).

AI Agent is an intelligent entity that can perceive the environment, make decisions and perform actions. Different from traditional AI, AI Agent has the ability to gradually complete a given goal by thinking independently and calling tools.

After the arrival of LLM, AI Agent was defined as an agent driven by LLM to realize automated processing of general problems.

We know that LLM is mainly good at processing and generating text. They can answer questions, write articles, generate creative content, help with programming, and more. But LLM is still a passive tool that only produces output when you give it input.

AI Agents provide a wider range of capabilities, especially in terms of interacting with the environment, proactive decision-making, and performing various tasks. It can be said that AI Agent is the key to truly unleashing the potential of LLM. It can provide powerful action capabilities for the core of LLM.

The main difference between AI Agent and large models is that the interaction between large models and humans is based on implementation. Whether the user is clear and unambiguous will affect the effect of the large model’s answer. There is no accurate and effective answer, not even the most capable ChatGPT.

The AI Agent only needs to be given a goal to work, and it can think independently and act on the goal. It will break down each planning step in detail according to the given task, and rely on feedback from the outside world and independent thinking to create for itself to achieve the goal.

For example, if you ask ChatGPT to buy a cup of coffee, the feedback given by ChatGPT is generally similar to “You can’t buy coffee, it’s just a text AI assistant.”

But you need to tell the AI Agent tool based on ChatGPT to buy a cup of coffee. It will first break down how to buy a cup of coffee for you and plan out several steps such as placing an order and payment on your behalf through an APP, and then follow these steps to call the APP to select takeout. Then call the payment program to place an order and pay. The process does not require humans to specify each step of the operation.

While both AI tools and agents are software programs designed to automate tasks, specific key characteristics distinguish AI agents from more complex AI software.

When an AI tool has the following characteristics, it can be considered an AI Agent:

**Autonomy: **AI virtual agents are able to perform tasks independently without human intervention or input.

**Perception: **Agent functions perceive and interpret their environment through various sensors (such as cameras or microphones).

**Reactivity: **AI agents can evaluate the environment and respond accordingly to achieve their goals.

**Reasoning and decision-making: **AI agents are intelligent tools that can analyze data and make decisions to achieve goals. They use reasoning techniques and algorithms to process information and take appropriate action.

Learning: They can learn and improve their performance through machine, deep and reinforcement learning elements and techniques.

**Communication: **AI agents can communicate with other agents or humans using different methods, such as understanding and responding to natural language, recognizing speech, and exchanging messages through text.

Goal-oriented: They aim to achieve specific goals, which can be predefined or learned through interaction with the environment.

**In terms of categories, AI agents can currently be divided into autonomous agents (Autonomous Agents) and generative agents (Generative Agents). **

Autonomous agents such as Auto-GPT can automatically perform tasks and achieve expected results based on people’s needs through natural language. In this cooperation model, the autonomous agent mainly serves humans and is more like an efficient tool.

Generative agents, such as the Westworld town jointly created by researchers at Stanford and Google or the humanoid robots in “Westworld”, live in the same environment, have their own memories and goals, and not only interact with humans, but also interact with each other. Other robot interactions.

Regarding AI agents, the 86-page LLM-based Agents review paper recently launched by the Fudan University Natural Language Processing Team (FudanNLP) comprehensively summarizes the current status of intelligent agents based on large-scale language models, including: the background, composition, and application of LLM-based Agents scene, and the much-discussed agency society.

Having said so much, many friends may still not have an intuitive feeling about AI agents. Don’t worry, below we will use a comparative case to deepen your understanding.

Ai intelligent agents penetrate into various fields

AiAgent.app is a web application that allows users to create custom AI agents to perform specific tasks and achieve goals.

Below, Wang Jiwei Channel will look at the advantages of AI agents through the comparative experience of using AI agents and directly using LLM.

For example, if you want to know the news and trends in the AI industry in the past month, enter in Claude: a summary of the latest news and trends in the AI industry in the past month.

The result obtained is as shown below:

As you can see, Claude only listed a few abstracts of news information related to AI.

Enter this paragraph in AiAgent.app, it will first break down your needs into ten tasks, then interact with the user through prompts to complete each task, and output the results for each task. Obviously, the content about the recent AI industry obtained in AiAgent.app is more comprehensive than that obtained by directly using other LLMs.

Is it possible to obtain this content directly using large models? Theoretically, it can be completed by entering more, but it needs to be entered at least ten times, and the accuracy of the input cannot be guaranteed, and sometimes you don’t even know what information you want to get.

In AiAgent.app, you only need to enter one sentence, and it will analyze your possible needs and list relatively comprehensive content goals, guiding you to accomplish what you want, and the efficiency will increase several times.

Comparing the two, it is clear that AI Agent is superior in terms of richness and efficiency of content acquisition. This kind of information content agent is of great value to media practitioners, industry analysts and other professions, and can greatly reduce the time to obtain research data.

There are now some such agents targeting more precise user groups and application scenarios. For example, GPT Researcher launched by Columbia University is an Agent for researchers based on ChatGPT, which can create various research reports for users to promote research.

This case is only about content acquisition. In fact, agents for multiple application scenarios have emerged, which are enough to mobilize more software applications and even hardware devices to complete various tasks.

For example, some people have used AutoGPT to order meals, book tickets, take taxis and shop; the 25 AI Agents in Stanford’s Westworld town are walking, dating, chatting, drinking coffee and sharing the news of the day every day; Google Deepmind has launched the use of Robotic agents for robotic arms to automatically perform various tasks; Amazon has also launched Amazon Bedrock Agents to automatically decompose enterprise AI application development tasks; IBM Watson Health has been helping doctors diagnose, treat and monitor patients in many hospitals.

Although Ai Agent has not been popular for a long time, it has been supported by many companies in many fields as soon as it appeared. The multi-model capabilities of large language models coupled with today’s greater computing power have allowed Agents, which were proposed many years ago, to quickly gain value and be implemented in more fields with super penetration rates.

With the emergence of open source AI Agents such as MetaGPT, more technology providers and entrepreneurial teams have introduced Agents, and more organizations have recognized and accepted Agents. It will inevitably quickly become the main model for LLM implementation in various fields, helping thousands of industries to change their business. Good application LLM.

Inventory of 60 AI Agents around the world

AiAgent.app mentioned in the above case is one of the representative AI Agent products that has been gaining momentum in recent months. Multiple agents at home and abroad, including this AI agent, can be seen in the project inventory list below.

In order to let everyone better understand the AI Agents that have been launched so far, Wang Jiwei Channel (id: jiwei1122) divides these AI Agents into media reports, domestically launched, industry-based, overseas others and GitHub projects. They will be gradually rewarded in the future. The project library classifies these Agents into different categories.

The AI Agents reviewed in this article include both AI Agents frameworks and tools, and AGENT products based on some open source frameworks. At the same time, most projects and products are autonomous agents.

Because some manufacturers are relatively low-key and do not publicize to the outside world, the AI Agents listed in this article are not complete, so it is also called the incomplete list of AI AGENTs. More manufacturers and entrepreneurs are welcome to contact Wang Jiwei’s channel after reading this article, and we can all contribute to the prosperity and development of the AI AGENT ecosystem.

AI Agent reported by media

1、Auto-GPT

Auto GPT is a free open source project on Github that combines GPT-4 and GPT-3.5 technologies to create complete projects through APIs.

Different from ChatGPT, users do not need to constantly ask questions to the AI to obtain corresponding answers. In AutoGPT, they only need to provide it with an AI name, description and five goals, and then AutoGPT can complete the project by itself. It can read and write files, browse the web, review the results of its own prompts, and combine them with said prompt history.

Auto-GPT is one of the first examples of GPT-4 operating fully autonomously, pushing the boundaries of what artificial intelligence can do.

2、AgentGPT

AgentGPT allows you to configure and deploy autonomous AI agents. Just name your custom AI and tell it to start any goal imaginable, and it will try to achieve it by thinking about a task to accomplish, performing the task, and learning from the results.

3、Baby AGI

This is an AI-driven task management system. The system uses OpenAI and Pinecone API to create, prioritize and execute tasks. Create tasks by analyzing the results of previous tasks and predefined goals, and use OpenAI’s natural language processing (NLP) and Chroma to store and retrieve task results in context.

The appeal of Baby AGI lies in its ability to autonomously solve tasks and maintain predefined goals based on the results of previous tasks, as well as effectively prioritize tasks.

4、Jarvis （HuggingGPT）

A unique collaboration system developed by Microsoft that can use multiple AI models to complete a given task, with ChatGPT acting as the task controller. The project, known as JARVIS on GitHub, is now available for trial on Huggingface (hence HuggingGPT), an agent that works extremely well with text, images, audio and even video.

The way it works is similar to how OpenAI demonstrates the multi-modal capabilities of GPT 4 through text and images, but JARVIS goes a step further and integrates various open source LLMs for images, videos, audio, etc., and can also connect to the Internet and access files. For example, you can enter a URL from a website and ask questions about it.

5、Aiagent.app

Ai Agent is a web application that allows users to create custom AI agents to perform specific tasks and achieve goals. AI agents work by breaking down goals into smaller tasks and completing them one by one. Benefits include the ability to run multiple AI agents simultaneously and democratizing access to cutting-edge technology.

AI Agent also boasts features such as inline code blocks with syntax highlighting, and seamless collaboration with third-party platforms. The tool is free to use and provides a simplified way to build AI agents without requiring more technical knowledge.

6、CamelAGI

Camel AGI is a generative AI tool that enables users to solve given tasks by role-playing autonomous AI agents. Of course, users need to enable Java to use this tool. Camel AGI allows users to complete tasks using AI agents and provides the option to log in with Google or star the tool on Github.

7. “Westworld” simulation Westworld town

For this project, researchers from Stanford University and Google created an interactive sandbox environment containing 25 generative AI agents that can simulate human behavior. They walked in the park, drank coffee in cafes, and shared news with colleagues, showing surprisingly good social behavior.

For example, starting from a user-specified concept that an agent wants to hold a Valentine’s Day party, the agent automatically spreads party invitations over the next two days, makes new friends, asks each other to go on dates and parties, and coordinates events at the right time. Time to show up at parties together.

8、GPT-Engineer

GPT-Engineer is an open source AI tool that allows users to specify what they want to build and then have a clarifying conversation with the AI to generate the required codebase. The tool is designed to provide a simple and flexible user experience, allowing users to adapt and extend its functionality according to their needs.

The tool includes functions such as specifying the identity of the AI agent, storing communication history with GPT4, and rerunning message logs. Contributions to the project are welcome, and interested individuals can refer to the roadmap, projects, and issues available on the GitHub repository. GPT-Engineer aims to be an open platform for developers to explore and build their code generation toolbox.

9、MetaGPT

MetaGPT, launched by Meta Corporation, is a multi-agent framework that uses single-line input to generate APIs, user stories, data structures, competitive analysis, etc. This framework can serve as product managers, software engineers, and architects. The framework can act as an entire software company, orchestrating SOPs with just a single line of code.

MetaGPT is integrated with human SOP process design. Therefore, LLM-based agents generate high-quality, diverse, structured documents and designs. MetaGPT is designed to make it easy to design solutions for complex tasks and provide problem-solving capabilities that are nearly comparable to human intelligence.

10、Amazon Bedrock Agents

Amazon Bedrock Agents released by Amazon allow developers to quickly create fully managed agents. By executing API calls to enterprise systems, Amazon Bedrock agents accelerate the release of generative AI applications that manage and execute activities.

Amazon Bedrock Agents simplify the rapid engineering and orchestration of user request tasks. Once set up, these agents can autonomously build prompts and securely enhance them with company-specific data to provide natural language responses to users. These advanced agents have the ability to infer the necessary actions to automatically handle user requests.

11、nvidia Voyager

Voyager, jointly launched by NVIDIA, California Institute of Technology, and others, uses GPT-4 to guide learning Minecraft agents through the pixel world. It should be noted that Voyager relies on code generation rather than reinforcement learning.

Voyager is the first life-long learning agent to play Minecraft. Unlike other Minecraft agents that use classic reinforcement learning techniques, Voyager uses GPT-4 to continuously improve itself. It does this by writing, improving, and transferring code stored in an external skill library.

This results in small programs that help with navigation, opening doors, mining resources, crafting pickaxes, or fighting zombies. GPT-4 unlocks a new paradigm in which “training” is the execution of code and “training models” are the skill code base that Voyager iteratively assembles.

12、RoboAgent

The joint research team of Meta and CMU took two years to successfully develop the RoboAgent universal robot agent. RoboAgent has achieved 12 different complex skills through training with only 7,500 trajectories, including baking, picking up items, serving tea, cleaning the kitchen and other tasks, and can be generalized and applied in 100 unknown scenarios.

RoboAgent stays on task no matter how much interference it encounters. The goal of this research is to establish an efficient robot learning paradigm that addresses the challenges of dataset and scene diversity. The researchers proposed the Multi-Task Action Blocking Transformer (MT-ACT) architecture to handle multi-modal multi-task robot datasets through semantic enhancement and efficient policy representation.

13、Inflection AI Pi

The core brain of the personal AI Agent product Pi launched by Inflection AI is the Inflection-1 large model developed by the company, and its performance is comparable to GPT-3.5. Unlike popular general-purpose chatbots, Pi can only carry out friendly conversations, offer concise advice, or even just listen.

Its main characteristics are compassion, humility, curiosity, humor and innovation, and good emotional intelligence. It can provide unlimited knowledge and companionship according to the unique interests and needs of users. Since Inflection developed Pi, it has been determined that Pi will serve as personal intelligence (Personal Intelligence), not just a tool to assist people in their work.

14、HyperWrite

Hyperwrite is an AI writing agent tool that helps creative writers of any level write faster and more confidently. It includes features like auto-write and type-ahead to generate original paragraphs and come up with ideas for overcoming writer’s block.

The tool is available as a free Chrome extension and can be used on any website without interrupting workflow. It is used and trusted by professionals, students, and creators around the world to increase their productivity.

15、GPT Researcher

GPT Researcher is an AI-based autonomous agent used to conduct comprehensive online research on a variety of tasks. Inspired by AutoGPT and the “Plan and Solve” prompt, the tool aims to improve the speed and determinism issues found in current language models, “delivering more stable performance and higher speeds by working in parallel agents rather than synchronously operating” .

According to the team, GPT researchers facilitate research by generating relevant research questions, aggregating data from more than 20 web sources, and leveraging GPT3.5-turbo-16 and GPT-4 to create comprehensive research reports.

AI Agent launched in China

After continuous exploration and experimentation, domestic AI agent-related products have also begun to emerge. Here are five products.

1. Alibaba Cloud ModelScopeGPT

Alibaba Cloud Mota community launched the first large-scale model calling tool in China, ModelScopeGPT. Through this tool, users can call other artificial intelligence models in the Mota community by sending instructions with one click, thereby realizing large and small applications. models work together to complete complex tasks.

ModelScopeGPT is based on the open source large language model (LLM) AI Agent (agent) development framework ModelScope-Agent. This is a general, customizable Agent framework for practical applications. It is based on open source large language models (LLMs) as the core and includes modules such as memory control and tool usage.

The open source LLM is mainly responsible for task planning, scheduling and reply generation; the memory control module mainly includes knowledge retrieval and (prompt word) management; the tool usage module includes tool library, tool retrieval and tool customization.

2. Really intelligent TARS-RPA-Agent

TARS-RPA-Agent, launched by Real Intelligence in the field of hyper-automation, is a hyper-automatic agent based on the “TARS+ISSUT (Intelligent Screen Semantic Understanding)” dual-mode engine, with a “brain” and “eyes, hands and feet”. It is a new RPA model product that can autonomously dismantle tasks, perceive the current environment, execute and provide feedback, and remember historical experience.

TARS-RPA-Agent adopts a technical framework based on the TARS large model and ISSUT smart screen semantic understanding. The technical framework is divided into two layers: the bottom layer is the TARS series of large models including general basic models and basic models of various vertical industries, and smart screen semantic understanding technology; the upper layer is the ultra-large model that relies on these two key technologies to complete comprehensive upgrades and transformations. Automation products.

The core LLM of TARS-RPA-Agent is Real Intelligence’s self-developed vertical “TARS” large model based on a general large model base. The TARS large model has excellent mainstream functions such as text generation, language understanding, knowledge question and answer, and logical reasoning. ability.

3. OmBot ohm intelligent agent

At the 2023 World Artificial Intelligence Conference, Lianhui Technology released the OmBot, an autonomous agent (Auto AI Agent) based on large model capabilities, and launched the first batch of applications for typical scene requirements.

Lianhui autonomous agent contains the four core capabilities of cognition, memory, thinking, and action. As an automatic and autonomous agent, it runs in a loop in the simplest form. At each iteration, they generate Self-directed instructions and operations. Therefore, it does not rely on humans to guide commands and is highly scalable.

4. Lanma Technology Ask XBot

The Agent platform “Ask XBot” built by Lanma Technology is divided into two layers: the first layer is expert empowerment. Experts define workflows and teach them to machines through drag, drop, drag and dialogue interaction, thereby assisting front-line employees to build Methodology for more efficient work; the second level is for employees to use Agent. Front-line employees can communicate with the Agent through natural language and issue instructions, allowing the Agent to assist in data analysis, information retrieval and other tasks.

The company plans to build Ask Customers can be served more efficiently and intelligently on the platform.

5、ChatDev

ChatDev, launched by a joint research team of Tsinghua University, Beijing University of Posts and Telecommunications, and Brown University, is a generative agent. It is a chat-based end-to-end software development framework that leverages large language models (LLMs) to facilitate effective communication and collaboration among multiple actors (“gpt3.5-turbo-16k” version of ChatGPT) in the software development process.

The main purpose of ChatDev is game development through chat. Users only need to propose ideas, and the entire process from design to testing is completed by AI, and the entire process only takes seven minutes to complete.

AI Agent products for different fields

Before LLM appeared, some companies were already studying the combination of traditional AI and Agent. Therefore, the implementation of AI Agentmt in various fields is much faster than everyone expected.

Below are representative Agent applications in several industry fields.

In the medical field, Agents can help diagnose, treat and monitor patients. IBM Watson Health is an AI agent that analyzes medical data to identify potential health problems and recommend treatment options.

In the financial field, Agents can analyze financial data, detect fraud and make investment recommendations. Charles Schwab uses an artificial intelligence agent called Intelligent Portfolio to create and manage investment portfolios based on clients’ investment goals.

In retail business scenarios, Agents can provide personalized recommendations, improve supply chain management, and enhance customer experience. Amazon’s Alexa is an AI agent that can recommend products, place orders and track shipments.

In manufacturing, Agent can optimize production processes, predict maintenance needs, and improve product quality. General Electric uses an AI agent called Predix to monitor machines in real time to predict and prevent equipment failures.

In the transportation field, autonomous AI Agents can assist with route planning, traffic management, and vehicle safety. Tesla’s Autopilot helps self-driving vehicles and helps drivers park, change lanes, and drive safely.

In the education industry, Agnet can provide a personalized learning experience, automate administrative tasks and analyze student performance. Pearson’s AI agent Aida can provide students with feedback and suggest personalized learning paths.

In agriculture, AI agents can optimize crop production, monitor soil quality and predict weather patterns. John Deere is using an AI agent called See&Spray to detect and locate weeds without affecting crops.

Other AGENT products have been launched overseas

1、Cognosys

Cognosys is a web-based AI agent designed to revolutionize productivity and simplify complex tasks, using the most advanced AI technology to enhance your daily life.

2、Doanythingmachine

Easily manage your tasks with a “do-it-all” machine where the user’s personal AI agent will prioritize and complete your tasks for you

3、alphakit

An intuitive platform for creating and managing teams of goal-driven autonomous AI agents, all from your phone Create and manage autoGPT AI agent teams. Just define your goals and Alphakit takes care of the rest.

4、GPTConsole

GPTConsole is a revolutionary command line interface (CLI) designed to give developers the advantages of artificial intelligence. It goes beyond traditional terminal functionality to enable users to perform complex tasks using prompts.

5、Finishes

Convert your knowledge base into AI chat in 2 minutes by providing a link to the knowledge base. Fini provides users with a tireless AI agent ready to answer customer questions immediately 24/7.

6、Spell

Spell is an autonomous AI agent based on GPT4 that can be applied to daily efficient work. Spell also has much-needed features to help you work smarter and learn to harness the power of generative AI to generate one or more innovative autonomous agents that will work to solve your problems.

7、Aomni

Aomni is an information retrieval AI agent that can find, extract and process any data on the Internet for you, enhancing your research work. Aomni can use a variety of tools to intelligently plan your queries to get final results, including a full web browser that allows it to access any information on the Internet without the need for an API.

Aomni’s query planner is based on the current state-of-the-art AutoGPT architecture, intelligently planning and updating each request to ensure source correctness and diversity.

8、Fine-Tuner.ai

With Fine-Tuner.ai, users can build complex, tailor-made AI agents without technical skills or coding, just input your data and ideas. More than a dozen professional AI agents can create precise Q&A, document search, process automation, etc. for users through uploaded real-time data such as PDF, CV, PPT, and URL.

9、SuperAGI

An open source autonomous AI framework that enables you to quickly and reliably develop and deploy useful autonomous agents, and an infrastructure for building, managing, and running autonomous agents.

10、Yellow.ai

Yellow.ai is the leading enterprise-grade conversational AI platform that powers dynamic AI agents in the enterprise, designed to deliver human-like interactions through its no-code/low-code platform to increase customer satisfaction and increase employee engagement. .

11、Godmode

Enables users to run AutoGPT in the browser. Godmode allows users to deploy multiple AI agents at the same time to complete tasks using AI, and users can also use their own OpenAI API keys.

12、E42

E42 is a cognitive process automation platform that allows enterprises to create multifunctional cognitive agents to automate various processes across functions. The cognitive-driven, no-code platform integrates seamlessly with users’ existing technology and processes to unlock the highest value across departments. Users can use E42 to build their own AI agents, such as AI analysts and AI recruiters across vertical industries.

13、Thankful

Thankful’s AI agents are trained and tailored to work within your existing help desk, easily resolving high-volume customer inquiries via email, chat, SMS, and in-app channels. With the ability to understand, connect, resolve, personalize and inform, ThankfulAI agents deliver human-like service experiences with machine-like speed and inherently scalable expertise.

14、Aktify

Use Aktify’s virtual AI agents to clone your sales team without increasing headcount. Aktify will handle an unlimited number of unresponsive leads at scale) and consistently bring customers ready to talk to your sales team’s doorstep, it’s more than just an SMS chatbot.

15、TeamSmart AI

Increase your productivity with one-click access to TeamSmart AI. Aggregate content, generate code, draft tweets, and more right in your browser. ChatGPT instantly opens with a click of the icon or keyboard shortcut, providing instant access to a library of quality tips without logging in.

16、BrainstormGPT

BrainstormGPT integrates multiple agents, LLM and automatic search to simplify topic to meeting report conversion. Custom topics, user-defined roles, autonomous discussions by agents, and reports output within 20 minutes are approximately equivalent to 300 searches, 10 hours of discussions, and 100,000 text analyses.

17、AgentRunner.Ai

AgentRunner.ai is an autonomous AI agent creation tool that leverages the power of GPT-4 to create and train fully autonomous agents. Allows users to set goals for their agents and let them decide how to achieve those goals without any technical knowledge or programming skills.

The tool offers features such as creating autonomous agents with unique personalities, running the agent to perform tasks or learn new skills, deciding what the agent can do, and integrating with OpenAI or Google Cloud accounts.

18、Stay

Gista helps businesses engage with website visitors and convert them into leads 24/7, and its main features include building AI conversion agents and AI sales agents. Using Gista, businesses can easily convert website visitors into leads and build email lists.

19、Agent4

One of Agent4’s key features is the ability to create AI-powered virtual agents that can answer questions, help book meetings, listen to voicemails and provide summaries.

You can easily create custom interactions for agents, allowing them to answer questions and handle a variety of tasks in your brand’s voice. You can also choose how agents respond to calls in real time and decide if and when you need to speak to someone.

20、Cometcore AI

Cometcore AI is an innovative platform that provides a versatile set of AI-driven tools to improve productivity and communication. With Cometcore you can make, code and automate cute agents.

21、personal-assistant

An AI agent designed to handle everything from booking flights to conducting in-depth research and everything in between.

AI Agent project on Github

1、OpenAGI

OpenAGI is an open source AGI research platform specifically designed to deliver complex multi-step tasks, accompanied by task-specific datasets, evaluation metrics and a variety of scalable models. OpenAGI formulates complex tasks as natural language queries as input to LLM. The LLM then selects, synthesizes and executes the models provided by OpenAGI to solve the task.

The project also proposed a task feedback reinforcement learning (RLTF) mechanism, which uses task solving results as feedback to improve LLM’s task solving capabilities. LLM is responsible for synthesizing various external models to solve complex tasks, while RLTF provides feedback to improve its task-solving capabilities, providing a feedback loop for self-improving AI. The paradigm of LLM operating various expert models to solve complex tasks is a promising approach to AGI.

2、Agent-LLM

Agent-LLM is an AI automation platform designed to power efficient AI instruction management across multiple providers.

The agent is equipped with adaptive memory, and this versatile solution offers a powerful plug-in system that supports a variety of commands, including web browsing. With growing support for numerous AI providers and models, Agent-LLM continues to evolve to enhance a variety of applications.

3、AutoGPT-Next-Web

This agent can deploy the well-designed AutoGPT-Next-Web Web UI on Vercel with one click, and deploy your private AutoGPT-Next-Web web application for free with one click. Based on AutoGPT-Next-Web, users can use Vercel to deploy for free with one click and build a personal AutoGPT website in 1 minute.

4、MiniGPT-4

This Agent can use advanced large language models to enhance visual language understanding.

5、Mini-AGI

Mini-AGI is the smallest general-purpose autonomous agent based on GPT3.5/4. It combines powerful prompts, a minimal set of tools and short-term memory (thought chaining), with data augmentation via vector storage to be added soon, to analyze stock prices, perform cybersecurity tests, create art and order pizza.

6、Teenage-AGI

Inspired by several Auto-GPT related projects (mainly BabyAGI) and the paper “Generative Agents: Interactive Simulation of Human Behavior”, this Python project uses OpenAI and Pinecone to provide memory for an AI agent and Allow it to “think” before taking action (outputting text).

7、FastGPT

FastGPT is a knowledge base question and answer system based on the LLM large language model, providing out-of-the-box data processing, model calling and other capabilities. At the same time, workflow can be orchestrated through Flow visualization to realize complex question and answer scenarios.

8、DemoGPT

With DemoGPT, you can quickly create a demo using only simple sentences.

9、LocalAGI

Locally run AGI projects based on LLMDA, ChatGLM and other models.

10. ai-town (game category)

The well-known investment institution a16z’s open source AI town is an MIT-licensed, deployable starter kit for building and customizing your own version of an AI town. This is a virtual town where AI characters live, chat, and socialize.

11, gptrpg (game category)

gptrpg This repository contains two things: a simple RPG-like environment for an LLM-enabled AI agent, and a simple AI agent that connects to the OpenAI API to exist in that environment.

12. SFighterAI (Game Category)

The project is an AI agent trained using deep reinforcement learning to defeat the final boss in the game Street Fighter II: Special Champion Edition. The AI agent makes decisions based solely on the RGB pixel values of the game screen. In the provided save state, the agent achieves a 100% win rate in the first round of the final level.