Start free.
Pay when it's working.

Plans are priced monthly in BDT. Online payment runs through SSLCommerz and is still being finalised — our merchant account is in onboarding — so for now we set your plan up with you directly rather than charging online. Either way your subscription is controlled entirely by our backend: provisioning only happens after payment succeeds or a trial is approved, and plan limits are enforced server-side.

The backend enforces every limit. Automation pauses for the metered resource you've exhausted while your history and settings stay intact — nothing is deleted. You'll see it coming in your usage dashboard and can upgrade at any time.

Each inbound or outbound message the agent handles, on any channel you've connected — and each comment it replies to. We meter messages, model tokens, retrieval calls, and tool calls separately so you can see exactly where usage goes.

The feature is built, metered, and included on every plan — but Meta has to grant this app public Page access before a business can connect its own Page, and that approval is not in place yet. Until it is, we connect Pages by hand, so tell us and we will set yours up. Instagram comments are built too and stay switched off until Meta's Instagram permission is available to us; Facebook is what runs today.

Yes, over MCP. Every plan includes knowledge-base ingestion (PDFs, FAQs, product catalogs, web pages) and lets you connect your own systems as external data sources — paid plans simply allow more of them. Whatever your MCP server exposes, the agent can call, behind the same policy gate.

A full operational workspace. Your team gets a shared inbox with assignment and private notes, saved replies, labels, automation rules, satisfaction surveys, campaigns, a help centre, contact records, and reports — plus channels beyond Meta: a web chat widget for your site, email, Telegram, and SMS. You run the agent, knowledge base, usage, and billing from our dashboard, and step into the workspace in one click.

Completely. You write its system prompt, choose its model, set how adventurous it is, define the words that force a handoff to a human, decide how hard it tries before giving up, set its language, and pick the team it escalates to. There are no vendor defaults you can't override.

An AI call agent — a separate voice service on the same grounded, auditable core, answering from your knowledge base and handing the call to a person when it's out of its depth. And direct connectors for order lookup, customer lookup, bookings, and ticketing, which today you can already wire up yourself over MCP. None of these has shipped, and none is billed on any plan.

Strictly. Every tenant has its own knowledge base, tools, and audit trail, enforced at the database level. The agent only ever loads context scoped to your business — content is never merged across tenants.