Costs

Ask what your bots actually cost.

They answer the question a dashboard cannot: not what did I spend, but which bot, model or ad is costing more than it is worth.

Things you can just ask

Real prompts, not pseudo-code. Claude picks the tools itself.

“What did each bot cost me this month, and what is the cost per reply?

“Which model is eating the most credits, and would a cheaper one do for those questions?

“How many AI replies do I have left this month?

“Compare this month’s spend to last month and tell me what changed.

The 7 tools behind it

Tool	Needs	What it does
`get_bot_costs`	read	Spend per bot, with cost per reply and the model breakdown.
`get_credit_balance`	read	Credits and AI replies left this month, per workspace.
`get_ad_results`	read	Chats, leads, sales and spend per Click-to-WhatsApp ad, with the cost of each.
`get_meta_conversions`	read	Whether sales from ads reach Meta, and which ad conversations were reported as a Purchase.
`set_meta_purchase_stage`	write	Mark a pipeline stage as a sale, so ad leads who reach it are sent to Meta as a Purchase.
`set_meta_tracking`	write	Connect a WhatsApp number to the Meta dataset, so ad leads and sales are reported to Meta.
`list_feedback`	read	Bug reports your team has sent from inside the console — what broke, who hit it, and which are still open.

read works with any key. write is refused unless the key allows it — the server checks before the tool runs, so a read-only key genuinely cannot change anything.

What it will not do

Worth knowing before you wire it into something that matters.

Both tools are read-only. Nothing here can change a plan or buy credits — that is deliberate.
Costs are what the provider charged, with no markup, so the numbers will not match a competitor’s "per resolution" figure.
Replies your team sends by hand are free and do not appear as cost.

Also over MCP

Manage bots

Six tools cover the whole lifecycle of an agent: find it, read it, change it, and check the change before a customer sees it.

Reply to customers

The shared inbox and the sales board. The important sequence is take_over → send_reply → hand_to_ai: without the first call the AI will answer over the top of you.

Add knowledge

The one that earns its keep is delete_knowledge: when a bot gives a wrong answer, the fix is almost always removing a stale chunk, not adding a new one on top.

Connect your store

The point of sync_catalog is that the bot quotes today’s price and today’s stock, not a snapshot from setup day — and templates are the only way to reach someone after the 24-hour window has closed.

Start free — 300 credits, no card

Ready when you are

Your first agent, answering in five minutes.

Create your bot — free Talk to the demo bot →