Tencent has released teamai-cli, an open-source command-line tool designed to solve a growing problem in engineering organizations: how to standardize AI coding agent configurations across teams without forcing everyone onto a single vendor's platform.
The tool, available on GitHub and npm, provides a Git-based workflow for managing and distributing AI agent resources — skills, rules, documentation, environment variables, hooks, and MCP server configurations — across a team's members. An administrator pushes configuration to a shared repository; team members pull it, and their local installations of Claude Code, Codex, Cursor, CodeBuddy, OpenCode, and other supported agents automatically receive the updates.
The core workflow is deliberately simple: teamai push creates a branch and opens a Merge Request for review. Once merged, teamai pull (triggered automatically by a SessionStart hook) syncs the latest resources to each member's local AI tools. The system supports both project-scope and user-scope installations.
What makes teamai-cli notable is its breadth of agent support. The tool works with Claude Code, OpenAI Codex, Cursor, Tencent's own CodeBuddy, WorkBuddy, OpenCode, OpenClaw, Hermes, DeepSeek Harness, and Qoder. Each agent's skill directory gets populated with the team's approved configurations — Claude Code skills go to ~/.claude/skills/, Codex to ~/.codex/skills/, and so on.
Beyond basic configuration distribution, teamai-cli includes several advanced features. A role-based distribution system lets administrators define which skills each team member receives based on their role. Tag-based filtering provides more granular control. A knowledge base system, powered by BM25 search and a code graph built from AST analysis, lets agents automatically search accumulated team knowledge before tackling tasks.
The tool also includes a session friction detection mechanism: when an AI session involves significant user corrections, tool call retries, or other friction signals, teamai-cli suggests sharing those learnings with the team. Over time, this creates a feedback loop where hard-won debugging insights propagate automatically.
For organizations running multiple AI coding tools — increasingly common as teams experiment with different agents for different tasks — teamai-cli addresses the configuration management nightmare that results. Without it, each team member must manually configure each agent, leading to drift and inconsistency. With it, a single push from an admin propagates standardized configurations to every tool in the stack.
The project is MIT-licensed and actively maintained, with contributions welcome. It represents a growing category of infrastructure tools emerging to support the operational needs of teams deploying AI coding agents at scale.
