Skip to main content
Your audit found the opportunity. This recipe turns it into working software your client can click, in the same conversation, with no code. It works by connecting two tools to one AI assistant: the Audity connector (your audit data and deliverables) and the Lovable connector (an AI app builder that deploys live apps). The assistant pulls the opportunity from Audity, writes a build brief, and drives Lovable to ship a demo at a real URL.
Lovable’s agent connection is currently a research preview on their side, so tool names and behavior may change. This recipe was verified against Lovable’s official MCP server (mcp.lovable.dev) as of July 2026. You need a Lovable account (free tier works; builds spend Lovable credits from your account, not Audity credits).

What you’ll end up with

  • A deployed proof-of-concept app at a *.lovable.app URL demonstrating one audit opportunity (an intake portal, an approval workflow, a dashboard, whatever the opportunity describes)
  • Grounded in the real numbers from your audit: the client’s pain points, current process, and ROI figures
  • Ready to screen-share in your next client call, or iterate further in Lovable’s editor

Prerequisites

1

Connect Audity to Claude

If you haven’t already: npx @auditynow/connect, or follow Connect Claude. Verify with “List my Audity projects.”
2

Connect Lovable to Claude

In Claude’s Settings → Connectors, add Lovable’s connector (https://mcp.lovable.dev). Lovable uses OAuth: a browser window opens, you sign in to your Lovable account, and you’re done. See Lovable’s own docs for the current setup flow.
3

Have a project with completed analysis

The recipe pulls from your opportunities and ROI calculations, so run the audit first. See Run a full audit workflow.

The recipe

1

Pick the opportunity

In a Claude chat with both connectors active:
Claude calls audity_get_project_opportunities and audity_get_roi_calculations, then recommends. Pick one.
2

Generate the build brief

Reviewing the brief before the build is the step that makes the output client-presentable instead of generic. Edit it in plain English until it matches what you’d actually pitch.
3

Build and deploy

Claude drives Lovable’s builder (create_project, then deploy). First builds typically take a few minutes. You get back a live *.lovable.app URL.
4

Iterate

Look at the deployed app, then refine in the same chat:
You can also open the project at lovable.dev and continue in their editor directly.

A worked example prompt

Everything above in one message, for when you know exactly what you want:

Tips

  • Always review the brief before building. The demo is only as sharp as the brief; thirty seconds of editing beats a regenerate.
  • POC means POC. Position it to the client as a directional demo, not production software. It has sample data and no real integrations.
  • Data stays illustrative. Have Claude use realistic sample data, not real client records. The Lovable app is hosted outside Audity.
  • Costs land on your Lovable plan. Builds spend Lovable credits; nothing in this recipe spends Audity credits.
  • Keep the audit as the anchor. The strongest close pairs the live demo with the deliverables it came from: the final report and deck carry the numbers, the POC makes them tangible.

Troubleshooting

Check Settings → Connectors and confirm both Audity and Lovable show as connected in the current chat. Some clients require enabling connectors per-conversation.
Lovable’s connection uses OAuth sessions that can expire. Re-authorize the Lovable connector and retry; your Audity connection is unaffected.
Skipping the brief-review step is the usual cause. Regenerate the brief, add the specific workflow and metric that made the opportunity compelling, and rebuild.
Your token may lack scopes or your plan may not include the operation. See Connect Claude → Troubleshooting.

What’s next