Try it with your agent
Paste one prompt into your coding agent and get a campaign on staging.
Have an idea for a campaign, such as a referral race or a launch quiz? Copy the prompt below into Claude, ChatGPT, Gemini, or another AI agent. It sets up Domino, signs you in, asks what you want to achieve, and builds a first version on a staging site you can open and share.
You don't write any code. You answer the agent's questions, approve one sign-in, and confirm its plan.
I want to try Domino, a platform for community campaigns with quests, points, rewards, referrals, and leaderboards. Set it up from the command line, learn what it can do, ask me what I want to achieve, and build a first version on a staging site I can open. Work through the steps in order. Where a step says to wait for me, end your turn.
Step 1. Get somewhere you can run commands. You need a shell on my computer with Node.js 24, Git, and pnpm. If pnpm is missing, run corepack enable pnpm; if that fails, use npx --yes pnpm@VERSION wherever this prompt says pnpm, with the version from the project's packageManager field. If you can't run commands here, for example because this is the ChatGPT, Claude, or Gemini chat app, stop and tell me where to continue. Recommend the coding agent from the same company: Codex for ChatGPT, Claude Code for Claude, Gemini CLI for Gemini. Cursor, GitHub Copilot in VS Code, or any other agent that runs terminal commands also works. Give me short install steps, tell me to open it in a new empty folder and paste this prompt again, then wait. The latest version of this prompt is at https://console.domino.run/docs/try-with-your-agent.txt.
Step 2. Install the Domino CLI. Run npx -y @domino-sdk/relay-cli@latest --version to check that it works. Until a project is checked out, run every Domino command as npx -y @domino-sdk/relay-cli@latest followed by the command. Inside a checked-out project, use pnpm exec domino instead, so the CLI matches the project.
Step 3. Sign me in. Run npx -y @domino-sdk/relay-cli@latest login. It prints a code and a sign-in link. Send me the code first, then the link, in one message, and wait for the CLI to confirm that I approved it. If I don't have a Domino account yet, I can sign up from that link. Never ask me to paste a token or password into this chat.
Step 4. Explore before you ask me anything. Run whoami --json to see my organization and the projects I already have. For an existing project, run quests list --project PROJECT_ID --environment test --json to see what it contains. Read https://console.domino.run/docs/agents and the index at https://console.domino.run/docs/llms.txt, then run capabilities --json to see what Domino supports today.
Step 5. Interview me. In a few short questions, ask what I want to achieve: my next real campaign or launch, who it's for, what they should do, why they would take part, and how I'll know it worked. Ask for my brand name and website if I have them. Suggest options based on what you learned, and if I already have projects, ask whether to build on one of them or start fresh. Then wait for my answers.
Step 6. Propose a plan. Describe the campaign, the quests, points, rewards, leaderboards, or referrals it uses, and how it will look. For a new project, pick a starter: campaign, classic, waitlist, referral, discord, loyalty, or api. Say plainly if part of my idea isn't supported yet. Wait for me to confirm or change the plan.
Step 7. Build it on staging. For a new project, run create PROJECT_NAME --template STARTER. For an existing project, run checkout PROJECT_ID. Work inside the directory it creates, install dependencies with the exact command it prints, then read AGENTS.md, CAMPAIGN.md, and the Domino skill at node_modules/@domino-sdk/relay-cli/cli/skills/domino/SKILL.md and follow them. Save the brief we agreed on in CAMPAIGN.md. Build the plan and run pnpm exec domino check and pnpm test. A project you just created is yours alone, so commit and push to main. An existing project may be shared: work on a branch and follow the source and release workflow in the skill's hosted reference (references/hosted.md), which merges the latest main before you push the integrated result to main. Then run pnpm exec domino stage --commit FULL_COMMIT_SHA --json for the commit you pushed to main and wait for the build to finish.
Step 8. Hand it over. Send me the staging link, what to try first, what you tested, and any decisions still open. Everything runs on test data. Don't publish to live; going live needs my review in Domino Console. If a Domino command fails or something is confusing, report it with domino feedback as the Domino skill describes.
The same text is at /docs/try-with-your-agent.txt, so you can send someone the link instead of the full prompt.
What happens
Your agent moves to a terminal
The agent needs to run commands on your computer. A chat app can't, so the agent tells you which coding agent to switch to, such as Codex for ChatGPT, Claude Code for Claude, or Gemini CLI for Gemini. Open it in an empty folder and paste the prompt again.
You approve the sign-in
The agent sends you a code and a sign-in link. Open the link, sign in or create an account, and enter the code. The agent picks up from there. You don't need Console for the rest.
The code works for 15 minutes, and the sign-in page shows how many minutes are left. If it expires, ask the agent for a new one.
The sign-in page also shows when the request was made, the country it came from and the client that sent it. Approve only a code your own agent or terminal just asked for: whoever holds an approved code acts as you.
You answer a few questions
The agent looks at your organization's projects and at what Domino supports, then asks about your campaign: who it's for, what they should do, and why they would take part. It proposes a plan, says plainly if part of your idea isn't supported yet, and waits for you to confirm it.
You get a staging link
The agent creates a project from a starter, or checks out one you already have, builds the plan, and stages it. You get the link, what to try first, and any decisions still open.
Staging runs on test data, so you can try it without touching real members. Nothing is published to live until you review it in Console.