Build intelligent agents that think, act, and work together. Automate complex workflows and turn your ideas into systems that run for you.
Start with one workflow, prove it works, then expand to the next.
Write down what you want to automate and what a good result looks like.
Set up your agent, add instructions, and connect the tools it needs.
Run the workflow, read the execution history, and refine what to change.
Pick a template and adapt it. Each one opens in the agent builder.
Press run to watch each step change state, then open the real builder.
Track status, cost and approvals. Figures on this page are sample data.
| Task | Agent | Model | Status | Cost |
|---|---|---|---|---|
| Enrich new account | Prospect Researcher | Gemini | Success | $0.02 |
| Prepare call brief | Call Brief Builder | GPT | Success | $0.03 |
| Draft renewal reply | Follow-up Drafter | Claude | Needs approval | $0.04 |
| Classify support ticket | Ticket Triage | Gemini | Running | — |
| Summarize weekly data | Report Summarizer | Claude | Success | $0.11 |
Give agents more responsibility as you gain confidence in them.
The agent drafts. You decide what happens next.
The agent acts after a person approves the step.
Workflows start from a trigger you run by hand and report their results. Scheduled triggers are not available yet.
Compare runs and refine them. On the roadmap.
Every capability is labeled with its real status.
Bring in the apps your team already uses. Each card shows its real connection state.
Read how it fits together, or go straight to the canvas.
Lunaire is free while it is in preview. Paid plans will be announced before anything changes.
Everything that is live today, for you or your team to try.
For groups that build and review agents together.
For organizations that need access control and audit trails. Runs in your browser for now.
The short version of what works today and what does not yet.
Yes. While Lunaire is in preview, everything that is live is free to use. No prices have been set for later plans, and the page will say so before that changes.
Your agents, workflows, knowledge and run history are saved in this browser. When the page is opened on claude.ai while signed in, they also sync to your own private storage for this page, so they follow you across devices. “Clear all data” in Settings wipes both.
When the page is opened on claude.ai while signed in, AI steps run through Claude with no API key. If you host the page yourself, you can point it at your own proxy in Settings and use your own key.
Integration cards show their real state. Right now they are setup states only, so a Tool step reports “Integration not connected” until the service is connected on a server.
You decide. Add a Human Approval step before anything important and the run waits for a reviewer to confirm or reject it.
Single sign-on and real multi-user accounts, which both need a server. Knowledge sources, agent teams, usage tracking, roles and audit logs now work, but they are stored in this browser only, so each person sees their own data. Use Export and Import on the Teams page to share work.
Build your first AI agent with Lunaire. Free while in preview.
Get Started