TLDR: An 18-year-old student from the University of Port Harcourt (UNIPORT), Obinna Chimdi, has developed ChatATP, an AI agent that enables large language models (LLMs) like ChatGPT, Gemini, and Claude to perform direct actions on websites. The platform, built over six months, aims to bridge the gap where LLMs can chat but not execute tasks end-to-end. ChatATP utilizes custom toolkits and a unique ‘Agents2 protocol’ to facilitate interaction between AI and web functionalities, although it currently faces challenges in user adoption and funding amidst a competitive AI landscape.
Port Harcourt, Nigeria – Obinna Chimdi, an 18-year-old student of Mathematics and Computer Science at the University of Port Harcourt (UNIPORT), has unveiled ChatATP, an innovative AI agent designed to empower large language models (LLMs) to execute tasks directly on websites. The development addresses a significant limitation of current LLMs, which, despite their conversational prowess, often fall short in performing end-to-end actions such as booking flights or sending emails.
Chimdi’s journey to creating ChatATP began with a personal frustration. He observed that while tools like ChatGPT are powerful, they are not inherently built to solve problems comprehensively. ‘If you ask them to analyse data and send the report to someone by email, for example, they stop halfway. You still need to copy, paste, or do the rest yourself,’ Chimdi explained to Techpoint Africa. This realization spurred him to combine different AI models and integrate execution layers, culminating in ChatATP – an AI that not only understands but also acts.
Building ChatATP was a six-month endeavor for Chimdi, who started programming at the age of 16 using his father’s mobile phone. A self-taught backend developer, he had to acquire frontend development and UI/UX design skills due to a lack of collaborators. His motivation was deeply rooted in his admiration for Mark Zuckerberg and a passion for automation, aspiring to build his own AI company.
At its core, ChatATP functions by creating ‘toolkits’ that act as a translation layer between a website and an LLM. These purpose-built connectors expose specific actions, allowing AI to perform tasks like checking flight availability, sending emails via Gmail, creating contacts in HubSpot, or querying university timetables. Each toolkit is meticulously defined with its required inputs, expected outputs, and safety protocols.
The centerpiece of ChatATP’s architecture is the ‘Agents2 protocol,’ which Chimdi describes as ‘like HTTP for AI agents.’ This universal protocol standardizes how AI models and toolkits communicate, aiming to enhance interoperability and scalability in the burgeoning AI agent ecosystem.
Despite its innovative approach, ChatATP is currently a developer-centric tool, requiring users to configure API keys from LLM providers, install domain-specific toolkits, and select their preferred AI model. So far, Chimdi has onboarded only three users: his neighbor and two course mates. The platform faces an uphill battle for wider adoption, constrained by the need for significant compute resources and funding.
The competitive landscape for AI agents is fierce, with established open-source projects like AutoGPT and LangGraph offering similar functionalities for autonomous agents and complex workflows. Newer entrants such as Lumio AI provide multi-model workspaces, allowing users to switch between various LLMs like GPT-4, Claude, Gemini, and Mistral. These competitors often benefit from robust developer communities and substantial financial backing, resources that ChatATP currently lacks.
Also Read:
- AI’s Transformative Role in Nigerian Education: A Tribute to Professor Fafunwa’s Enduring Legacy
- Paper2Agent Initiative Transforms Scientific Research into Interactive AI Agents
However, the unique ‘Agents2 protocol’ could serve as ChatATP’s key differentiator, offering a level of flexibility in integrating multiple LLMs and toolkits that many rivals do not emphasize. To truly establish itself, Chimdi must demonstrate ChatATP’s scalability and attract users beyond his immediate network, or risk it becoming another experimental project overshadowed by better-funded and faster-moving players in the rapidly evolving AI domain.
