Agent Teams AI: Multi-Agent Orchestration with Kanban Board

What changed

The Agent Teams AI project, a TypeScript-based framework for multi-agent systems, has recently seen its latest release, version v2.17.1, on September 28, 2026. This release signifies continued development and support for the project, which aims to allow users to act as 'bosses' to AI agents.

Why it matters for builders

This project is designed for developers looking to create sophisticated AI applications that go beyond single-model interactions. It facilitates agent-to-agent communication and collaborative task completion, managed through a user-friendly Kanban board. The architecture supports a vast array of LLM providers and models, including Codex, Claude, OpenCode, Cursor, Grok, GitHub Copilot, Kiro, Z.AI, Xiaomi, MiniMax, and Kimi, with support for over 300 models and 200 LLM providers, including free models that require no authentication.

Key Features:

• Agent Orchestration: Agents handle tasks autonomously, communicate with each other, and review each other's work.
• Kanban Board Interface: Provides a visual overview of agent progress and allows for high-level commands.
• Broad Model Support: Integrates with numerous LLM providers and models, offering flexibility.
• Company Building: Enables the construction of AI companies with multiple specialized teams.

Practical impact

Developers can leverage Agent Teams AI to build and manage complex AI workflows. The project's focus on agent autonomy and inter-agent communication, coupled with a visual management tool, can streamline the development of AI-powered products and services. The extensive model support means builders can select the most suitable and cost-effective LLMs for their specific needs.

Caveats and source limits

The provided repository metadata indicates a recent release and a strong developer signal count, suggesting active development. However, specific details regarding the implementation of agent communication protocols, the exact capabilities of the Kanban board, or performance benchmarks are not detailed in the summary. The project is licensed under AGPL-3.0, which may have implications for commercial use. Further investigation into the project's documentation and code would be necessary to fully understand its operational intricacies and limitations.