HoneyCold, B2B prospecting without cold guesses
From Google Maps to a prepared first call.
See the live product (opens in a new tab)Project at a glance
- The need
- Prepare outreach with useful information, then track conversations in one tool.
- My contribution
- Product design and development: company search, AI analysis, call preparation and sales follow-up.
What was built
- Search and audit
- Call scripts and emails
- CRM and reminders
- FR / EN versions
Public product with a no-account demo. The evidence describes how it was built; no commercial outcome is claimed.
- Nature
- Online product (SaaS) · AI-assisted B2B prospecting
- Period
- Since March 2026
- Client
- HoneyColdOnline product, designed and run by me
The problem
Phone prospecting too often means calling blind: copying listings from Google Maps, guessing who to call first, improvising what to say. And B2B cold calling is regulated: a prospect’s objection must be honoured, everywhere and for good.
The approach
One tool, from search to a prepared call. It starts from public sources (Google Maps, company registers), compares what it finds with the user’s offer and prepares a documented first call. Three structuring choices: the product can be tried without an account, through a public demo; AI credits are deducted on the server, atomically; people’s rights (objection, erasure) are product features, not a legal page.
The solution
- Search by trade and city, enriched with public company data (SIRET, INSEE).
- Two-level AI audit: "Éclair", quick, and "Expert", which analyses the website in depth.
- A call script and email tailored to each prospect, ready to copy in one gesture.
- Built-in CRM: statuses, scheduled reminders, CSV export, dashboard.
- Team mode, and the product is available in French and English.
- Objection and suppression: a shared suppression list, rights requests tracked with a thirty-day deadline.
What is not claimed
No usage figure and no commercial result is claimed here. The figures below describe how the product is built, not how successful it is.
Evidence
Every figure carries its source. No commercial result is claimed.
1 737
Automated tests passing
SourceProject log, state of 1 October 2026
2 104
FR/EN translation keys checked, zero gaps
SourceProject log, audit of 1 October 2026
Technical details
The technologies used for this project. Features and their limits are detailed in the narrative.
- Next.js
- React
- TypeScript
- Supabase
- Stripe
- Claude
- Playwright
- Redis
Expertise involved
Gallery
The home page and its no-account demo. 
Before / with: five tasks of the caller. 
The product’s features, one card per use. 
Same page, on a phone.
A similar need?
Let's discuss your context and what you would like to simplify.