- TeamAI treats AI agent skills, rules, MCP tool configs and model settings as Git-managed assets with branch-and-merge review workflows.
- The tool supports 16 AI coding agents including Tencent's own WorkBuddy and CodeBuddy alongside Claude Code, Cursor and Codex.
- Internal testing across 13 development tasks showed 76% lower model costs and 6.1% higher quality scores when cheaper models used shared team knowledge.
- TeamAI has been integrated into Tencent Cloud's ClawPro agent management platform and has already reached GitHub Trending.
Git-Native Agent Collaboration
Tencent Cloud on October 10 released TeamAI as an open-source project under the MIT license, positioning it as a shared foundation for how teams work with AI agents. The tool stores agent skills, working specifications, tool configurations and project knowledge in a standard Git repository, allowing teams to review, update and version-control these resources using familiar merge-request workflows.
When a team member submits a new skill, TeamAI automatically creates a branch and opens a merge request for review. Once merged, agents that support session-start hooks automatically pull the latest updates at the beginning of each new conversation, eliminating manual synchronization.
Broad Agent Compatibility and Cost Savings
According to Tencent Cloud, TeamAI currently supports 16 AI coding agents, including its own WorkBuddy and CodeBuddy products as well as third-party tools such as Claude Code, Codex, Cursor and GitHub Copilot CLI. The shared resource scope covers skills, rules, MCP server configurations, sub-agents and, more recently, unified model configuration distribution.
Tencent Cloud reported that in an internal test involving 13 development tasks, integrating shared team knowledge into lower-cost models improved comprehensive quality scores by 6.1% while reducing model invocation costs by 76%. The company has also integrated TeamAI into ClawPro, its intelligent agent management platform.
Sources & context
Go to the original material. Company claims remain attributed to their sources.
01


