Credentials. Meta access settings are entered in the dashboard, so keep the token limited to the pixel and restrict who on your team can change it.
Who it is forA merchant running Meta ads who sees many orders refused at the door and wants ads tuned to delivered sales.
The problemMeta Pixel fires on checkout submit, so ads are optimized for orders that may be refused at delivery.
The outcomeMeta receives a Purchase event at confirmed delivery, with a test stream available to verify events before going live.
What you need first
- Node.js 22.12+ and npm
- Wrangler CLI (npm install -g wrangler)
- A free Cloudflare account
- Carrier and Meta credentials if you connect delivery or ads (optional)
Steps
- In the dashboard, open the Meta Pixel and CAPI configuration and enter your pixel details.
- Choose Purchase at confirmed delivery as the conversion event.
- Turn on Test Mode so CAPI events go to Meta's test stream (test_event_code) while you verify them.
- Review the CAPI event audit log (capi_event_log) to confirm each send attempt and its result, then turn Test Mode off.
Prompt for your agent
In the CodFlow repo at https://github.com/bighadj22/codflow, check the Meta CAPI configuration for store <store name> with Purchase set at delivery. Expected output: the current conversion event, the Test Mode state and the last five entries in the CAPI event log. Verify that Test Mode is on before any live traffic is sent.
Caveat. The README says a production deployment with Worker-to-Worker fetch needs a custom domain for cod-server, so verify your CAPI sends after that is in place.
Use this in your AI agent
Ask your agent
Using the Deduped MCP, get the setup steps for bighadj22/codflow and "Optimize Meta ads for confirmed deliveries, not form fills"
Add Deduped to Claude Code once
claude mcp add --transport http deduped https://dedupedai.com/mcp
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeShare: X LinkedIn Hacker News Email Copy link
Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer