Skip to content

> my aiYour team's AI, set up once

One organization manifest gives every AI harness generated guidance, mounts, launch profiles, and local operating context — on any machine, with one command.

Install

sh
curl -sSL https://raw.githubusercontent.com/fluxinc/my-cli/master/install.sh | sh

Run my update to update to the latest release; re-running the installer also works.

First Run

sh
my init acme --name "Acme"
my onboarding
my ai codex

my onboarding launches guided onboarding in a harness when run interactively. Use my onboarding --no-agent for the deterministic setup walkthrough. my ai codex performs the same root resolution and guidance freshness check before starting a harness. my init creates a private manifest repo (the control plane) plus a content repo at ~/acme/workspace (the actual workspace), all local and working offline; my publish later creates the private remotes and pushes both.

The Operating Shape

~/acme/
├── .my-cli/          # workspace identity and local state
├── workspace/      # manifest-declared content mount (its own repo)
├── repos/          # opted-in catalog repositories
├── personal/       # local-only scratch
├── .mcp.json       # generated local MCP config
├── AGENTS.md       # generated root guidance
└── CLAUDE.md       # compatibility pointer when supported

The organization manifest lives in its own private repository outside the umbrella — the workspace is a mount of things the manifest defines, and day-to-day work never edits the manifest itself.

Part of a Toolchain

my is the organization layer of a broader agentic stack: org context and knowledge for every agent and human, from one manifest. It composes with gnit (git-native multi-repo workspaces, the umbrella's publish substrate) and clawdapus (governed agent containers whose cognition is mediated by the cllama proxy) — manifest roles compile into contained fleet agents that carry the my CLI as a governed work surface. Gated organization services (credential brokers, human-reviewed communications) are declared in the manifest and consumed the same way by human and AI operators.

Start with the quickstart, then read the model for the boundary between the public CLI and a private organization manifest.

Released under the MIT License.