UNIVERSAL OPERATIONAL CLI /// HUMAN + AGENT

Make the
safe thing
obvious.

Omni turns service APIs into a command line whose action level is visible before anything can happen. Two live integrations: Trello and Tailscale.

OMNI // COMMAND VIEW
$ omni observe tailscale device list
  → Lists device IDs, hostnames, OS, and last-seen time.

$ omni observe trello board list
  → Lists boards. No mutation. Safe for unattended approval.

$ omni create trello card create LIST_ID --name "Ship it"
  → Creates one card. Interactive approval required.

$ omni --policy read-only delete trello card delete CARD_ID
  × blocked by read-only policy

Two services.
Live today.

Not a speculative catalog. These are native integrations with real setup flows, capability discovery, and fixed reviewed API requests.

SERVICE / 01

Trello

Boards, lists, and cards—read first, then create, move, archive, or delete with a visible action level.

omni describe trello

SERVICE / 02

Tailscale

Tailnet devices, routes, users, and policy with compact agent-oriented observation and deliberate administration paths.

omni describe tailscale
OBSERVE

Read is first-class.

Safe operations begin with observe, so tools and people can understand their effect at a glance.

ACT

No surprise flags.

Create, move, archive, delete, and authorize each get their own left-edge command path.

RESTRICT

Policy travels local.

Read-only and no-delete modes are enforced by Omni, independently of an agent allowlist.

One binary.
Zero ceremony.

Linux x86_64 / ARM64 and macOS Intel / Apple Silicon. Downloads the matching checksum-verified release to ~/.local/bin; no sudo, package manager, or account required. It reports install, upgrade, or already-current status and refuses to replace another program named omni.

curl -fsSL https://toppk.github.io/omni/install.sh | sh

Prefer to inspect first? Read install.sh · Download the binary directly, then chmod +x omni_linux_amd64 (choose the matching asset).