Stop Setting Up Your Team’s AI Tools Over and Over. A Look at Tencent’s TeamAI

A new member joins the project. You want them to start using an AI agent right away, but the questions keep coming. “Which skills do we use in this repository?” “Where are the review rules?” “Do I need to configure MCP servers too?” You explain everything individually, send over configuration files, and notify them again whenever something changes. As the team grows, this work quietly adds up.

Tencent’s “teamai-cli” is an open-source CLI for managing a team’s AI configurations and resources together in a Git repository. It supports sharing skills, rules, agents, hooks, MCP servers, model settings, and more. It promises a setup in which members configure it once, then receive updates published by administrators with each session.

For teams where only the AI enthusiasts have well-developed setups, or everyone has their own way of doing things, it is a tool worth a look. Here, we explore what TeamAI can do and the steps to understand before adopting it.

Sharing AI Configurations Through Git#

The idea behind TeamAI is straightforward: manage the team’s AI resources in a shared repository. That repository holds more than configuration files. It can also contain skills, rules, agents, hooks, MCP configurations, models, and more.

A “team’s way of working” might include:

• Rules listing what to check during code reviews
• Skills for carrying out specific tasks
• Roles and settings for individual agents
• Connection information for AI tools, such as MCP servers

Instead of having each member copy settings manually, the team uses a Git repository as the source of updates. When someone improves a skill, they add the changes to the shared repository, and other members receive the update. Using Git also lets teams review configuration changes and keep a history, just as they do with code.

Several AI tools are supported. The GitHub listing names Claude Code, Codex, Cursor, GitHub Copilot CLI, OpenCode, and others, and includes a table showing support for each feature. However, not every tool supports every feature. Check the compatibility table in the official repository for the combination of agent and features you want to use.

Getting Started as an Administrator#

Installation uses npm, the Node.js package manager. First, install the CLI.

npm install -g teamai-cli

Next, create a shared repository on the Git host your team uses, such as GitHub or GitLab. The official documentation lists GitHub, GitLab, GitCode, CNB, TGit, and private Git services. The intended setup allows members to push branches while keeping the default branch protected.

Once the repository is ready, the administrator initializes it as follows:

teamai init https://github.com/your-org/your-repo

If you do not have a shared repository yet, the documentation also suggests using a template from teamai-hub. These templates include ready-to-use skills, rules, and review agents. A practical approach is to check whether they fit your team’s needs before bringing them into your own repository.

Members Choose Where to Initialize#

When setting up TeamAI, members choose where to place its resources. The default is workspace scope: move into the project directory, then run the command.

cd /path/to/my-workspace
teamai init https://github.com/your-org/your-repo

To use the settings at the user level across multiple workspaces, specify --scope user.

teamai init https://github.com/your-org/your-repo --scope user

For team use, it helps to agree on the appropriate scope at the outset. The choice depends on whether the settings belong to a specific project or should follow a member across multiple workspaces. Before establishing operating rules, check where the settings apply and how they interact with existing configurations in your actual environment.

After setup, the system is designed to automatically retrieve administrator-published updates to skills, rules, and other resources with each AI session. This reduces the need for members to run synchronization commands themselves. During adoption, check the finer details, such as when updates take effect and how configuration changes appear during ongoing work.

Three Layers: Execution, Context, and Improvement#

TeamAI’s product overview divides its features into three broad layers.

The first, “Team Execution,” covers features that tell AI how the team works. It includes skills, rules, environment settings, agents, hooks, MCP, and models. Starting here lets you try standardizing work procedures and review criteria.

“Team Context” is the layer intended to help AI understand the team’s background. It lists learnings, codebase graphs, and teamwiki. This area is presented as a beta. It is best approached as something to explore while checking how code structure and team-specific knowledge are added and how they influence responses.

“Team Improvement,” also in beta, covers usage and sessions. Its aim is to turn the results of AI use into improvements for the team, but what information is collected and who can see it are operational concerns. Before using it, consult the relevant official documentation rather than relying solely on the public overview.

Check Compatibility Feature by Feature#

When reviewing the list of supported tools, do not judge by product names alone. For each agent, the table shows support for skills, rules, environment settings, agents, hooks, MCP, models, learnings, codebase, teamwiki, usage, and sessions.

In other words, asking “Can I use TeamAI with this AI tool?” is not enough. First decide what you want to share—MCP settings, rules, or learnings, for example—and check the corresponding column in the compatibility table. The table also includes notes explaining that some environment settings are available only through teamai env exec, and that some rules apply without path restrictions.

The compatibility table may change as versions and development progress. Recheck the latest information when adopting TeamAI, especially if you plan to use beta features or depend on a particular AI tool.

Decisions to Make Before Team Adoption#

Adoption requires more than distributing the CLI; the team also needs a process for managing the shared repository. Who can change skills and rules? How will changes be reviewed? How should user-level and project-level settings be used? If these questions remain unresolved, centralizing AI configurations can actually make management more complicated.

Starting small is a practical approach. Choose one repository and share just a few existing review procedures or skills. Have both an administrator and a member run initialization, then confirm that settings land in the intended locations and updates are retrieved. If everything works, gradually expand the resources you share.

TeamAI is released under the MIT license and welcomes discussions, issues, and pull requests on GitHub. To check its behavior and supported capabilities, you can also consult documentation such as Getting Started, Usage Guide, Git Providers, and Windows Setup.

For teams looking to reduce the work of redistributing configuration files for each AI tool, TeamAI looks like a useful starting point. Choose an agent you already use and one configuration you want to share, check the official compatibility table, and try initializing it with a shared test repository.

Source: https://github.com/Tencent/teamai-cli