Install a rig

rig hub use <source>

Examples:

rig hub use hub:maria/market-researcher
rig hub use hub:maria/market-researcher@1.2.0
rig hub use github:acme/agents-research
rig hub use ./downloads/market-researcher-1.0.0.rig.tgz
rig hub use registry:maria/market-researcher

The full grammar is on Source specifiers.

What it does

1. Resolve — work out where the artifact lives based on the source prefix.
2. Download — fetch the .rig.tgz to a temp file. For Hub installs, the CLI reads X-Rig-SHA256 from the response and verifies the hash.
3. Inspect — open the tarball, read rig.toml, validate the manifest.
4. Extract — write files into the target directory.
5. Record — write .rig/instance.toml with the source ref and install time so rig hub update knows where this came from.

What it doesn't do: install MCP servers, write env values, run shell hooks, or grant permissions. Those happen in the separate rig install step so they're explicit.

Flags

After install

cd <name>
rig install        # set up env, dirs, permissions
rig doctor         # confirm readiness

rig install is split out because it's the side-effecting step — see Run setup for the full picture.

Sandboxed test installs

To try a rig without polluting your workspace:

rig hub use hub:author/name --dir /tmp/test-rig
cd /tmp/test-rig && rig install && rig doctor

Nothing in /tmp survives a reboot, so this is a clean way to evaluate before committing.

See also

• Inspect before installing — read the manifest without extracting.
• Source specifiers — every supported source.
• Update — pull a newer version while keeping your local state.