tomershahar

Team Foundry

Name: Team Foundry
Author: tomershahar

AI community

SOURCE https://github.com/tomershahar/team-foundry

Description

One shared context layer for Claude Code, Cursor, Gemini CLI, and Codex — so the whole team's AI tools build from the same outcomes, decisions, and quality bar

Installation

This entry records only its repository, not the path inside it, so there is no exact command to give. Open the source below and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

README

team-foundry

**Your coding agents know the code. They do not know which customer matters, what the team decided, or what quality bar applies — so each teammate's AI gives different answers to the same question. team-foundry turns that missing product context into one shared, versioned layer that Claude Code, Cursor, Gemini CLI, and Codex all read from the repo.**

[![npm](https://img.shields.io/npm/v/create-team-foundry)](https://www.npmjs.com/package/create-team-foundry) [![npm downloads](https://img.shields.io/npm/dw/create-team-foundry)](https://www.npmjs.com/package/create-team-foundry) [![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](LICENSE)

![Three teammates - PM on Claude Code, Engineer on Cursor, Designer on Gemini CLI - connected to a shared AGENTS.md](assets/hero.svg)

Try a populated fictional team in 10 seconds, without changing your repo:

npx create-team-foundry playground

Or set it up for your own team:

npx create-team-foundry

Local-first and MIT licensed — no backend, no API keys, no telemetry. Each tool's loading route is verified in the [compatibility matrix](docs/compatibility.md).

![20-second demo: playground scaffolds a populated example team, then doctor scores its context health](assets/demo.gif)

Controlled comparison

On July 2, 2026, two fresh Codex background agents received the **same prompt**, runner configuration, and repository snapshot of Clearline — the fictional team the playground generates. Both ran on the runner's default model (the runner does not report which); neither case used a model override, so the comparison is like-for-like. The control could not read team-foundry context; the other could. We fixed the five-point scoring rubric before running either case.