teamai-cli

Product Introduction

1. Definition: TeamAI-CLI is a command-line interface (CLI) tool and team-native AI orchestration platform designed to standardize and scale AI adoption across software development teams. It functions as a centralized "Harness" for managing AI agents, skills, rules, and team knowledge.
2. Core Value Proposition: It solves the problem of fragmented AI tool usage by providing a unified, git-based workflow to distribute and synchronize AI best practices, ensuring every team member and every AI agent (like Claude Code, Cursor, or CodeBuddy) works consistently with the team's accumulated knowledge and approved methodologies.

Main Features

1. Team Execution Harness: A git-centric distribution system for team AI resources. How it works: Team admins maintain a central repository containing skills (SKILL.md files), rules (.md files), environment configurations, Model Context Protocol (MCP) servers, and hooks. Team members run teamai init and teamai pull to sync these resources locally, where they are automatically injected into supported AI tools via hooks, ensuring uniform agent behavior.
2. Team Context (Knowledge Recall): An intelligent retrieval system that allows AI agents to automatically search and reference accumulated team knowledge before tackling a task. How it works: It uses a BM25 search algorithm boosted by a codebase knowledge graph. When enabled via teamai recall enable, it deploys a teamai-recall subagent. This agent performs a relevance pre-check, extracts keywords from the user's task, searches the teamwiki/ knowledge base (containing shared learnings and codebase graphs), and returns a structured summary to the main AI session.
3. Team Improvement (Friction-Based Learning): An automated system for capturing and sharing valuable team experiences. How it works: A session Stop hook analyzes "friction" signals like user interruptions, denied tool calls, or AI retries. High-friction sessions trigger a suggestion to run /teamai-share-learnings, which summarizes the session and pushes a learning document to the team repo. The teamai digest and teamai dashboard commands provide analytics on usage, interventions, and knowledge base health.
4. Codebase Knowledge Graph Import: Automatically parses source code repositories into a structured, searchable graph. How it works: The teamai import and teamai codebase --extract commands use a WebAssembly (WASM) tree-sitter parser for AST-based analysis (TypeScript, Python, Go) to resolve precise import/export and implementation edges, complemented by a heuristic regex track for other languages. This graph enables structurally-aware code search and navigation during knowledge recall.

Problems Solved

1. Pain Point: Inconsistent and siloed AI adoption within development teams, where each developer uses AI tools (e.g., GitHub Copilot, Cursor, Claude Code) with their own prompts and techniques, leading to variable code quality and repeated problem-solving.
2. Target Audience: Engineering leaders and development team managers seeking to enforce best practices; development teams of all sizes wanting to standardize AI-assisted workflows; open-source project maintainers aiming to onboard contributors with predefined AI skills.
3. Use Cases: Enforcing company-specific code review rules via AI agents; distributing troubleshooting playbooks for common production issues; synchronizing MCP server configurations for tools like PostgreSQL or Figma across an entire team; capturing and reusing solutions from difficult debugging sessions; ensuring new hires immediately benefit from the team's accumulated AI expertise.

Unique Advantages

1. Differentiation: Unlike individual AI coding assistants, TeamAI-CLI is a team-coordination layer. It doesn't replace tools like Claude Code or Cursor but orchestrates them, turning individual AI capabilities into a standardized, team-wide competency. Compared to manual wiki documentation, it actively injects knowledge into the AI's context.
2. Key Innovation: The friction-based, automated learning capture system. Instead of relying on manual documentation, it uses quantitative session signals (interruptions, retries) to identify and prompt the sharing of tacit knowledge precisely when it is most valuable, creating a self-improving loop for team intelligence.

Frequently Asked Questions (FAQ)

1. How does TeamAI-CLI integrate with existing AI coding assistants like GitHub Copilot or Cursor? TeamAI-CLI uses provider-specific hooks and configuration files (like CLAUDE.md or agents/ directories) that are automatically injected into the local workspace when you run teamai pull. It does not require modifying the AI tools themselves; it configures the context they operate within.
2. Is TeamAI-CLI suitable for a solo developer or a single user? Yes, a solo developer can use TeamAI-CLI as a personal knowledge management system for AI. You can initialize it with a template repository to manage your own skills, rules, and learnings across different projects and AI agents, creating a consistent personal workflow.
3. What version control systems (Git hosts) does TeamAI-CLI support? It supports GitHub, GitLab, GitCode, Coding.net (CNB), Tencent Cloud DevOps (TGit), and any other private Git service that uses standard Git protocols, making it adaptable to enterprise environments.
4. How does the knowledge recall feature impact AI performance or latency? The recall subagent first runs a fast relevance pre-check (teamai recall --check). If the task is unrelated to team knowledge, retrieval is skipped entirely, minimizing latency. The search itself is optimized and runs locally.
5. Can I control which skills or rules are synced to my machine? Yes, TeamAI-CLI offers granular distribution controls. Admins can define roles and tags for skills/rules. Individual members can use teamai skill exclude to block specific skills and use teamai source to subscribe only to relevant external skill repositories.