Quickstart: install a rig

You'll need the CLI installed.

1. Find a rig

Browse the Hub in your browser, or search from the CLI:

rig hub search market-researcher

You'll get a list of author/name pairs. Pick one.

2. Install it

rig hub use hub:maria/market-researcher

This pulls the artifact, verifies its SHA-256, and extracts it into a new folder named market-researcher in the current directory. To install somewhere else:

rig hub use hub:maria/market-researcher --dir ./projects/market-research

You can install from places other than the Hub — see Source specifiers.

3. Set it up

cd market-researcher
rig install

rig install is the convergence step. It:

• Creates any folders the rig expects (data dirs, output dirs).
• Asks you for any required env vars and writes them to the file the rig declared.
• Merges the rig's declared Claude permissions into .claude/settings.json.
• Spawns the sync daemon if this is a shared rig.

It does not install MCP servers, run shell hooks, or install plugins. Those need explicit confirmation — see Run setup.

4. Check readiness

rig doctor

You'll see a list of checks (env vars, tools, MCP servers, directories) and their status. Anything red is a gap you need to fix before the agent can do its job. Most gaps come with a remediation hint — a shell command or a file path to look at.

If you run Claude Code inside the workspace, it'll read the same rig doctor output automatically and offer to fix gaps. See Agents.

5. Start working

claude

(Or whatever runtime the rig declared in [run].) The agent now has the prompts, MCP servers, permissions, and data it needs. Have fun.

When things go wrong

• rig doctor shows red checks — see Troubleshoot.
• MCP server fails to start — see [mcp].
• Need a different version — rig hub update --check and Update.