AI configuration that scales with your team — not just with you.

Stop re-explaining your context to your AI. Build it once, share it with your whole team.

Open source · Works with Claude Code & Gemini CLI

The AI productivity gap is a team problem.

Everyone configures alone

Each developer wires up their own system prompts and personas. The same setup work happens N times — and N times, slightly differently.

Context walks out the door

Personas, system prompts, and hard-won workflows live in someone's dotfiles. When that engineer leaves, so does the institutional knowledge their AI was carrying.

Your AI is a stranger to your codebase

The model has never seen your conventions, your modules, or the trade-offs your team already settled three quarters ago. Every session re-litigates the basics.

Power without coordination is noise.

Individual AI tooling is zero-sum. Ten developers each iterating on their own system prompts produces ten versions of the same wheel — and zero compounding gains. The teams pulling ahead aren't the ones with smarter models. They're the ones with a shared layer underneath. Like a git history for system prompts: everyone contributes, everyone inherits.

Three layers. One coherent stack.

System prompt layers

Layered system prompts

Personal profile, team norms, expertise, and seniority — structured into a coherent context your AI loads automatically. Configurable per person, shared across your team.

Community zone

Shared skills and agents

A collaborative sandbox where your team builds and shares reusable skills, custom commands, and agent personas — written once, installed anywhere.

Harness loop

The framework learns from use

When an agent hits friction, it tags it. A curator clusters the signals, files issues, and ships fixes back into the shared config. Every workflow papercut becomes a permanent improvement.

Six steps. Then it compounds.

1

Fork

Clone the CrewRig template into your team's GitHub org.
2

Profile

Drop in your team's stack, conventions, and review rituals.
3

Deploy

One script compiles the config into Claude Code, Gemini CLI, and every supported harness.
4

Share

Push to main. Every developer's next session inherits the update.
5

Improve

Agents report friction; the curator turns signals into fixes.
∞

Compound

Every friction your team tags becomes a ticket. Automatically.

Built for teams that ship.

Layered context, no conflicts

Personal, community, and project rules merge in a predictable order. Nobody's personal preferences ever leak into the team config — and the team config never overrides someone's local setup.

Reusable skills, on tap

A growing library of community-authored agent skills your team can import and customize. The capability you needed last week is probably already written.

Friction becomes fuel

A built-in feedback loop captures the rough edges agents encounter in real work, curates them, and ships fixes back to the shared layer. Your config gets sharper every sprint.

Tool-agnostic by design

Write the config once; ship it to Claude Code and Gemini CLI today, with more harnesses landing. Switching tools no longer means rewriting your system prompt library.

Secrets stay where they belong

Credentials, tokens, and private data never enter the shared layer or travel to the model. Least-privilege defaults, vault-friendly, GDPR-aware out of the box.

Up and running in minutes.

No accounts, no SaaS, no waiting list.

Step 1 — Clone the repo

git clone https://github.com/crewrig/crewrig.git

Get a local copy of the framework.

Step 2 — Install prerequisites

Read the README → Prerequisites and install the required tools:
Task · Gemini CLI or Claude Code · fzf · uv · yq

OS-specific install commands are in the README.

Your AI harness:

Step 3 — Initialize

claude /init-personal-profile
claude /init-soul

Build your personal profile and customize the agent identity.

Step 4 — Setup

task setup-claude-interactive

Deploys the shared config to your Claude Code harness.

View on GitHub →