Changelog
What shipped, when it shipped, nothing else.
Mastak launched in early access in June 2026. This page tracks what we shipped — not what we plan to ship. No roadmap promises, no vaporware. If it's listed here, it works.
- New
Telegram bot inbox — connect a Telegram bot, same AI sales pipeline as Instagram: Welcome → Qualify → Value → Objection → Close.
- New
Instagram comment-to-DM automation via the official Meta Graph API. Dev Mode approval, no waitlists for our users.
- New
AI sales pipeline with a 5-stage state machine. Each stage has its own prompt context; the bot knows where in the funnel the lead is.
- New
OpenRouter LLM routing — pick GPT-4o mini (default), Gemini Flash-Lite, or bring your own Claude key. Switchable per account from the UI.
- New
Multi-tenant architecture. Every account is row-level isolated — no data bleed between tenants.
- New
Flat-rate billing: FREE (rule-based, no AI), START ($12/mo, 1,000 AI conv), PRO ($24/mo, 3,000 AI conv + LLM choice), AGENCY ($59/mo, 10,000 AI conv + BYOK). 14-day trial — 300 AI conversations included.
- New
RAG knowledge base — paste your product info, the bot learns it and cites it in conversations. Per-tenant pgvector embeddings, no Pinecone.
- Improved
Human-in-the-loop inbox. Operator takes over a lead, AI pauses for that conversation. Configurable auto-resume timeout.
- New
Early codebase: webhook receiver, BullMQ job queue, and the comment trigger engine that starts DM conversations.
- New
pgvector RAG pipeline with per-tenant embedding namespaces. Postgres-native, no external vector DB.
- New
Lead scoring and funnel state engine. Every lead has a score, a stage, and a full conversation history.