Turn your contact form into a conversation.
4.8(6)27TechnologyStartupsMarketing
Static contact forms convert badly, and the submissions that do arrive land in a shared inbox with no context. ConvoScore replaces the form with a chat widget, a floating bubble or an inline panel, that talks to the visitor like a person would. Google Gemini pulls structured details out of the conversation as it happens: name, company, role, need, urgency. A deterministic scoring function turns that into a priority the moment the chat ends, so your team knows who to call back first. It is a real self-serve SaaS, not a single-site demo: anyone registers, gets their own chatbot and dashboard, and drops one script tag onto their site to go live, no key to paste and no per-page config. Edits happen in a full-screen visual Studio and publish in under a minute with no redeploy. Rate limiting, a honeypot, a minimum-turn gate, and an LLM spam check keep the bot clean without ever showing a visitor a CAPTCHA.
Conversational data capture
No rigid script. A natural back-and-forth in 8 languages pulls out name, company, role, need, and urgency as structured data every turn, filling several fields at once when a visitor volunteers them.
Deterministic request scoring
A pure function, not an LLM's own "confidence" number: slot completeness (40 points), urgency (20), company size (20), and reason clarity (20), bucketed into High (75+), Medium (40 to 74), or Low. Same inputs always produce the same score, and every priority call traces back to the stored fields.
Self-serve multi-tenant SaaS
Any business registers, adds a site, and gets its own isolated chatbot, request pipeline, and dashboard, with no manual provisioning. A three-step sign-up gets a working bot in minutes. A superadmin console oversees every tenant site.
Full-screen visual Studio
Edit the bot's persona, greeting, quick replies, question flow, placement rules, and theme by clicking directly on a live preview of the conversation. The panel you edit is the preview. A classic form view is there for bulk edits. Publish a new version with no code change and no redeploy; it goes live within a minute.
One keyless script tag
A single `<script src="https://www.convoscore.com/widget.js" async>` adds ConvoScore to any site. It resolves which site and which chatbot to show from the page it runs on, so the same tag goes on every page and nothing needs a per-site key. It renders as a floating bubble or as an inline component pinned to a spot on the page (support form, pricing page, footer), and survives client-side navigation on single-page-app sites. The dashboard previews the widget against the business's real live domain before they paste anything.
Request inbox
A filterable, searchable "Requests" table, a detail drawer with the full transcript and the score breakdown, and a status pipeline that saves the moment you click. An email hits your inbox the second a High-priority request comes in.
Analytics
Daily activity, a conversion funnel, a recent-activity feed, and a per-chatbot breakdown, all in the dashboard.
Abuse defense without a CAPTCHA
IP-hashed rate limiting (raw IPs never touch the database), a honeypot field, a minimum-turn gate before a conversation can convert, and an LLM-driven spam and prompt-injection classifier. Four layers, and real visitors never feel any of them.
Error observability built in
Every unhandled error, server and client, lands in one deduplicated table with a superadmin view; the first sighting of a new error type emails the operator once.
Multilingual end to end
The marketing site, dashboard, chatbot persona, system prompts, and email notifications all localize across 8 languages (Arabic, Dutch, English, French, German, Portuguese, Spanish, Turkish) through a single config.
83 people supporting this launch
4.8
6 ratings
Full-Stack Developer with 3+ years of production experience. Shipped customer solutions at Insider, launched three platforms of my own, Linknown, Nivorai, and QuizCupid.
AI video with native sound, paid per clip. Nothing stored.
A calmer Mac menu bar, one click away.
Enterprise Software. Intelligent Automation. Global Impact.
Comments
27Full-Stack Developer with 3+ years of production experience. Shipped customer solutions at Insider, launched three platforms of my own, Linknown, Nivorai, and QuizCupid.
Built solo, full-stack, end to end: real multi-tenant auth and data isolation, a Gemini integration with schema-constrained structured output instead of prompt-and-pray text parsing, a deterministic scoring engine that is fully unit-tested and auditable, a full-screen Studio where the live chat preview is the editing surface, a keyless embed that resolves itself from the host domain and previews against a customer's real site before install, self-serve billing through Paddle as Merchant of Record (no company required to start), and a security model that never needs a CAPTCHA. It is live in production at convoscore.com, not a mockup: a running product with self-serve signup, real embeds, and real conversations.
Using a deterministic scoring function on top of the Gemini-extracted fields is a smart split — the team can actually audit why a lead got its priority. Can users tweak the scoring weights (e.g. weigh urgency higher than company size), or is it fixed per account?
Creator of Contractor OS — a job costing and profit tracking system for contractors.
I like the idea of replacing a static contact form with a conversation that also captures structured details like company, need, and urgency. The priority scoring after the chat is especially useful because it helps a team decide who to follow up with first. Does the scoring logic work out of the box, or can users customize it for different types of businesses?
Great products https://methcrystals.com/
Keeping the score breakdown with the transcript makes follow-up easier to explain. If Gemini times out or a visitor prefers a standard form, can the widget still capture their details and send the request without a score?
GuestNiche: SEO tools for audits, keyword research, link building and content planning. https://guestniche.com/
For agencies qualifying leads from different client sites, can a tenant export the captured fields, score breakdown and transcript as CSV or send them to a CRM? Showing the export format would help teams evaluate how it fits their follow-up workflow.
Very Very Very cool
Startup & SEO enthusiast
The deterministic scoring split is what stands out to me — a pure function over slot completeness, urgency, etc. keeps the score explainable, while Gemini only handles extraction. Too many AI lead-qualification tools let the model grade itself. The IP-hashed rate limiting plus honeypot instead of a CAPTCHA is also the right trade-off: every friction point you remove from a contact form measurably lifts completion. Curious how the minimum-turn gate behaves on low-intent visitors — does the bot shorten the flow when answers are thin, or always run the full question set?
Congrats on the launch! Swapping static forms for a chat that scores leads the moment it ends is smart - knowing who to call back first is the real pain. Nice call dropping CAPTCHA for a honeypot.
The abuse-defense design stands out here — IP-hashed rate limiting plus a honeypot and a minimum-turn gate gives real spam protection without forcing a CAPTCHA into the conversation, which is exactly the friction these widgets should avoid. Keeping the deterministic score breakdown next to the transcript in the inbox is also smart; it makes the score explainable to clients instead of a black-box number.
Seems interesting app
ProxyAI. One agent, every storefront, zero migration.
Showing the score breakdown helps teams audit lead priority. Consider making weights configurable so smaller companies with urgent needs are not consistently ranked below larger prospects.
Founder of BuyHub, a search engine for collectors and resellers across 100+ marketplaces.
Thanks for sharing, it's a great website
Founder of BuyHub, a search engine for collectors and resellers across 100+ marketplaces.
Thanks for sharing, it's a great website!
Founder of BuyHub, a search engine for collectors and resellers across 100+ marketplaces.
Thanks for sharing this product
Professional ICT Analyst
This is a smart upgrade from the traditional contact form. Turning a passive form into a conversation that captures context, qualifies leads, and automatically prioritizes follow-ups can save teams a lot of time while helping them focus on the prospects most likely to convert. What would make the biggest difference for your business: more leads, better-qualified leads, or faster follow-up?
Stop Losing Leads in WhatsApp. Start Converting Them with Realatic.
that's great to see your procudt
The page describes transparent scoring based on completeness, urgency, company size and reason clarity. Can a business change those weights for support requests, where urgency may matter more than company size?
The scoring angle is smart — most form tools stop at 'you got a lead' and never help you judge quality. One thing I'd watch: false negatives on legitimate but low-intent inquiries (e.g. a student asking a genuinely useful question). Does ConvoScore let you weight the rubric, or is the model fixed?
Your listing explains the scoring weights clearly. How do you handle company size or urgency when the visitor has not supplied that information? Showing “unknown” separately from a low value, with the transcript quote behind each extracted field and a manual correction option, would help teams avoid deprioritizing legitimate enquiries because of missing or misread answers.