One engine.
One YAML per client.
Zero fork-and-customise.
Conduit is the lead infrastructure that sits between your ad spend and your sales team. One codebase ingests, scores, qualifies, and follows up. A per-client configuration file controls every weight, threshold, template, and signal. No client-specific logic ever touches the engine.
Agencies run multiple clients on a single deployment, with strict data isolation per client_id. Operators run one client on their own infrastructure: their Supabase, their Twilio, their Meta keys, their bill. Onboarding a new client takes 3 to 6 hours.
Multi-source Lead Ingestion
Meta Lead Ads webhook, Gmail polling with AI email parsing, and CSV upload. Deduplicated by phone, normalised to a unified schema, idempotent against duplicate webhook deliveries.
LSQA Scoring Engine
Four-dimension priority scoring on every lead. Re-runs on every inbound message, so a cold lead can become hot mid-conversation. Every score change is written to history and is fully auditable.
AI Conversation Agent
A single Claude call replies, extracts new field values, detects buying signals, and detects opt-outs. Structured output, no parsing, no hallucinated formats. Reads your client knowledge base for grounded answers.
Follow-up Scheduler
Configurable sequences at any cadence. Template-first sends inside the WhatsApp 24-hour session window, free-form once the lead replies. Hot leads escalate immediately. Booked or lost leads cancel pending follow-ups automatically.
Operations Dashboard
Lead queue sorted by score with per-dimension LSQA breakdown bars. Snooze, reschedule, override stage, mark booked or lost, send free-form messages inside the active session window. Auth on Argon2id, password hash stored in config.
Per-client Data Isolation
Every table is partitioned on client_id with row-level security enabled. In owned-infrastructure deployments, the client owns the Supabase project, the Twilio account, the Meta credentials. You own the engine. They own the data.
Multi-client, Single Deployment
Run an entire book of clients on one Conduit instance. Each client gets their own YAML, their own knowledge base, their own templates, and their own scoring weights. Data is isolated by client_id with row-level security on every table.
Owned Infrastructure, One Tenant
Deploy Conduit on your own Railway or Render project. Your Supabase. Your Twilio. Your Meta credentials. Your Anthropic API key. The engine runs your logic and writes to your database. Nothing leaves your stack.
LSQA Model
Lead Signal Qualification & Allocation. Four dimensions. Configurable per client.
Every lead is scored across four signals: where they came from, how soon they need to act, how much they have told you, and how serious their budget is. Each signal carries a weight defined in the client config. Together they produce a single priority score from 0 to 100. The score updates on every reply.
AI disengages.
Hand-off to human.
Follow-up sequence runs.
Re-scored every reply.
Monitored for engagement.
Promotes on signal.
Auto-archive after 48 hours.
No further contact.
What Happens to Every Lead
From first webhook to closed deal, automatically.
Capture and Normalise
Meta form submission, Gmail notification, or CSV upload arrives. Raw payload stored. Idempotency key set so duplicate webhook deliveries are dropped. Phone-first deduplication. Fields normalised to the schema declared in client config. Initial LSQA score written.
Outreach Within Minutes
First WhatsApp message fires within 5 to 10 minutes of ingestion. Source-aware template selection: Meta leads get the simpler template, organic leads get the richer one. Template-first by default. Free-form is gated by the WhatsApp 24-hour session window.
AI Qualifies in the Thread
Every inbound reply runs through one Claude call that returns reply text, extracted fields, escalation flag, and disengage flag together. New fields are merged. The lead is re-scored. Conversation summaries are generated when history grows past the configured window.
Sequence Runs on its Own
APScheduler ticks every minute and sends due follow-ups. Skips silently if the lead already replied or changed status. Operators can snooze, reschedule, or cancel from the dashboard. Schedules are defined in client config, not hardcoded.
Hot Lead, Human Closes
When the score crosses the hot threshold or the AI detects a buying phrase from config, the agent disengages and the operator is notified. Mark the lead booked or lost from the dashboard, and pending follow-ups cancel automatically. Every state change is in scoring_history.
What You Actually Get
Operational specifics from the reference deployment.
Plans
Fully managed. India billed in ₹ via Razorpay; everywhere else in $ via Dodo Payments. Start free — no card required.
- Full + booking / catalog / Slack
- 5,000 AI conversations / mo
- 5 seats
- Full + voice / API
- 20,000 AI conversations / mo
- 15 seats
India prices are exclusive of 18% GST, added at checkout per the DPDP-compliant invoice.
International prices are billed by Dodo Payments as merchant of record, inclusive of destination tax.
Need volume, white-label, or agency terms? Talk to a human.
TALK TO
A HUMAN
FIRST
WHAT WE NEED FROM YOU
Your role: agency, operator, in-house team.
Your lead volume per month.
Your current sources and tooling.
WHAT HAPPENS NEXT
45-minute discovery call.
Sample client_config.yaml shared after the call.
Pilot in 1 to 2 weeks if there's a fit.
WHO WE WORK WITH
Performance agencies running multi-client lead-gen.
Operators in regulated verticals who need owned infrastructure.