ADK-TS Hackathon 2025
The ADK-TS Hackathon 2025 was a four-week virtual competition organized by IQ AI to encourage developers to build innovative artificial intelligence agents using the Agent Development Kit for TypeScript (ADK-TS). The event, which ran from September to October 2025, featured a $5,000 prize pool and attracted significant participation from the global developer community. [1]
개요
The ADK-TS Hackathon 2025 was a global, online competition designed to showcase the capabilities of IQ AI's open-source Agent Development Kit for TypeScript. The event officially began with its announcement and the opening of registration on September 9, 2025, and concluded with the submission deadline on October 7, 2025. The winners were subsequently announced on October 22, 2025. The hackathon saw substantial engagement, with 122 builders participating from around the world. This participation resulted in 50 unique project submissions, of which 34 qualified for the final round of judging. A total of $5,000 in prizes was distributed among the top 16 projects selected by the judges. [1]
Background and Goals
The hackathon was launched by IQ AI to accelerate the adoption and demonstrate the versatility of the ADK-TS framework. The event provided a hands-on opportunity for developers to explore the potential of building sophisticated, autonomous AI agents capable of solving complex, real-world problems. The organization outlined three primary goals for the competition. [1]
The main objectives were:
- Foster Innovation: To inspire developers to create novel applications and solutions using AI agents, pushing the boundaries of what was considered possible with the technology.
- Expand the Ecosystem: To encourage the creation of new tools, applications, and especially Model Context Protocol (MCP) servers. These expansions enrich the ADK-TS ecosystem by enabling agents to connect to and interact with a wider array of data sources and APIs.
- Grow the Community: To bring together a global community of developers, innovators, and builders passionate about AI agent technology, fostering an environment of collaboration and knowledge sharing. [1]
Tracks and Judging Criteria
To structure the competition and guide submissions, the hackathon was divided into three main tracks, each focusing on a different aspect of AI agent development. All submitted projects were evaluated by a panel of judges against a consistent set of criteria to ensure a fair and comprehensive assessment. [1]
Tracks
Submissions were categorized into the following three tracks:
- Best Agent Application: This track was designed for projects that demonstrated practical, real-world utility. The emphasis was on creating a tangible solution to a specific problem, showcasing the agent's effectiveness and user value.
- Best MCP Expansion: This track focused on projects that created new Model Context Protocol (MCP) servers. The goal was to expand the capabilities of the entire agent ecosystem by enabling AI agents to connect to and interact with new data sources or external APIs.
- Best Web3/Blockchain Use Case: This track was dedicated to agents designed to interact with decentralized technologies. Projects in this category included agents that interfaced with DeFi protocols, analyzed blockchain data, or managed smart contracts. [1]
Judging Criteria
Projects were evaluated based on four key criteria:
- Innovation: The originality and creativity of the project's concept and its approach to solving a problem.
- Technical Implementation: The quality, complexity, and overall execution of the code and system architecture.
- Real-World Utility: The project's potential for practical application, its impact, and its viability as a useful tool or service.
- Effective Use of ADK-TS: How well the project leveraged the core features, architecture, and capabilities of the ADK-TS framework in its implementation. [1]
Winners
The $5,000 prize pool was distributed among 16 winning teams. Three projects were selected as main track winners, receiving the top prizes, while ten other projects were recognized with bonus awards for their outstanding contributions in specific areas. [1]
Main Track Winners
The top three projects each received a $1,000 prize for their exceptional work in their respective categories. These projects were recognized for their comprehensive implementation, innovative concepts, and significant potential impact. [1]
| Category | Project Name | Description |
|---|---|---|
| Best Agent Application | CodeForge AI | A multi-agent development platform featuring over 10 specialized AI agents that collaborate within a unified workflow. CodeForge AI automates a wide range of software development tasks, including React code generation, security scanning, testing, and performance optimization. It also integrates with GitHub for commits, generates AI images for projects, and assists with documentation. The platform is accessible via web, Telegram, and voice commands. |
| Best MCP Expansion | OpsPilot | A Discord-native AI on-call team designed for DevOps automation. OpsPilot monitors Prometheus alerts, triages incidents, analyzes logs using semantic search, and can autonomously generate GitHub pull requests to fix identified issues. It utilizes multiple MCPs to connect various systems while ensuring human operators can oversee and approve critical decisions. |
| Best Web3/Blockchain Use Case | ChainInsight | An AI agent that streamlines the entire decentralized finance (DeFi) workflow within a single conversational interface. It performs research and analysis by accessing live Total Value Locked (TVL) data, identifies safe protocols, simulates transactions to assess risk, and prepares those transactions for signing with wallets like . |
Bonus Track and Noteworthy Winners
An additional ten projects were recognized for their outstanding contributions and innovative ideas. Each of these teams received a $200 award for their work. [1]
| Award / Project Name | Description |
|---|---|
| Most Practical Real-World Use Case: Confluent | An MCP server and AI agent that unifies Zoho Books and Xero accounting platforms. It allows users to manage financial data, create invoices, and automate notifications through Discord or a terminal interface. |
| Best Bot Integration: BingeBird | A Telegram bot for entertainment discovery. It fetches movie and TV data, generates review tweets based on the content, and posts them to X (formerly Twitter), using ADK-TS for state management. |
| Best Technical Implementation: Obrix | An AI-powered monitoring platform that detects wash trading and pump & dump schemes in real-time across Uniswap V4 pools. It achieves this by analyzing swap data with Large Language Model (LLM) agents. |
| Best Improvement to ADK-TS: Bazaaro | An integration of ADK-TS with Coinbase's CDP AgentKit. This project created an AI-powered service aggregator that uses a three-agent system to discover and orchestrate responses from multiple paid APIs. |
| Best Collaboration/Team Agent: On Chain Analysis Agent | A multi-agent system that delivers token intelligence. It coordinates specialized agents to detect new tokens, fetch market data, conduct web research, and synthesize the findings into actionable analysis for users. |
| Noteworthy Winner: Chain Pilot | An AI-powered blockchain terminal that uses natural language commands for multi-chain operations, including NFT portfolio management and transaction preparation. |
| Noteworthy Winner: BlockIQ | A blockchain analytics platform featuring custom MCP tools and smart caching. It transforms complex on-chain data into conversational insights and interactive visualizations. |
| Noteworthy Winner: ReflectIQ | An AI-powered journaling application with semantic search, automatic goal tracking, and team collaboration features designed for structured productivity. |
| Noteworthy Winner: KeeperDCA | An automated dollar-cost averaging (DCA) agent that utilizes the Delegation Toolkit. It allows users to set up recurring transactions with a one-time signature. |
| Noteworthy Winner: Mendel.AI | A drug discovery automation platform accessible via Telegram. It features six specialized agents that manage the entire Design-Make-Test-Analyze (DMTA) cycle to deliver ranked molecular candidates. |
