Prerequisites

Everything on this site is browsable without any local setup. You only need tooling when you want to act on what you find here - paste an infer agents add or infer skills install command, pull a card's OCI image, or open the "+ Add agent" PR.

The infer CLI

The CLI is what runs the copy-paste commands shown on every card. Pick one of the install methods below.

Install script (user-local)

The script defaults to /usr/local/bin, which needs root. Point it at a directory on your PATH that you own to avoid sudo:

mkdir -p "$HOME/.local/bin"
curl -fsSL https://raw.githubusercontent.com/inference-gateway/cli/main/install.sh \
  | bash -s -- --install-dir "$HOME/.local/bin"

Make sure $HOME/.local/bin is on your PATH. Pin a version in production with --version vX.Y.Z.

Nix flakes

The CLI ships a flake.nix exposing infer as both a package and an app.

Run it once without installing:

nix run github:inference-gateway/cli -- --version

Install it into your profile:

nix profile install github:inference-gateway/cli

Or add it to a system / home-manager config by referencing the flake input github:inference-gateway/cli and pulling packages.<system>.infer.

Flox

Flox can consume the same flake. Add infer to an existing Flox environment with:

flox install --flake github:inference-gateway/cli infer

Or declaratively, by editing the env's manifest.toml (flox edit):

[install]
infer.flake = "github:inference-gateway/cli"  # pin a tag in production, e.g. ".../v0.112.1"

Then flox activate puts infer on your PATH.

Verify

infer --version

Full installation options (container image, manual download with checksum verification, etc.) live in the CLI repo.

A running gateway

infer agents add writes an entry into your local config that points an A2A agent at an inference gateway. You need a gateway reachable from your CLI to actually invoke the agent - the registry itself never talks to a gateway, but the commands you paste expect one. See inference-gateway/inference-gateway for setup.

Network access

The site fetches both catalogs at runtime, and cards link out to source repos. You need outbound access to:

• cdn.jsdelivr.net - serves catalog.json for both agents and skills.
• raw.githubusercontent.com and github.com - reached when you open a source repo or the "+ Add agent" PR flow.
• ghcr.io - only when you later pull an agent's OCI image locally.

Optional: a container runtime

Docker or Podman is only required if you intend to run an agent container locally. Browsing and installing into the CLI config does not require one.