Build your agents by talking to one.
Every Hania workspace comes with an assistant that can do the work in your account. Tell it what you need and it builds the agent, connects the tools, adds your knowledge, tests the whole thing, and puts it live. It never sees your credentials, and it stops to ask before anything it can't undo.
Build it by asking
Say what the agent should do. It does the rest.
Describe the job in a sentence or two. The assistant creates the agent, writes the instructions, and picks the settings, using the same builder the dashboard uses. Nothing is hidden from you: every agent it makes is a normal agent you can open and edit by hand.
- From a sentence, or from a blueprint. Describe it in your own words, or start from one of the ready-made roles and let the assistant adapt it to your business.
- Your knowledge, added for you. Point it at a page on your site or paste the text in, and it attaches it to the right agent.
- Change your mind out loud. "Make it warmer." "Never quote prices." "Use a different voice." It edits what it already built and leaves everything else alone.
- Live when you are. It sets up the website widget and hands you the snippet, and it can put a voice agent on a phone number you already own. For WhatsApp, Messenger and the rest it walks you through the connect steps in the dashboard, and it can move an existing channel to a different agent.
Tools
It connects your tools. Even the ones we've never heard of.
Ask for the integration you want and the assistant sets it up, asks for the credential when one is needed, and tests the connection before you rely on it. If we don't have a ready-made connector, it can go and read the service's API documentation, try the calls, and build you a custom tool from what it learns.
- Ready-made connectors for the services teams use daily, set up and assigned to the right agent.
- Any HTTP API, researched and configured from the service's own docs.
- It reads the real schema. The assistant works from the actual configuration and settings definitions rather than guessing at field names, which is why what it builds works the first time.
- It tests what it makes, and shows you the call and the response, so you see it working before it goes anywhere near a customer.
Credentials
It sets up the tool. It never sees your key.
When a tool needs an API key or a token, the assistant doesn't ask for it in the chat. It opens a secure form, and the value goes straight into encrypted storage. It never appears in the conversation, never in what the model reads, and never in our logs.
It can't read the key back afterwards either. Neither can anyone else, including us. The assistant configures the tool around a credential it has never been shown, then tests the connection and tells you whether it worked. If a key stops working or you rotate it, it asks for the new one the same way.
Give each tool the narrowest credential that does the job. A read-only key where reading is all you need holds at the service itself, which is stronger than any setting in Hania.
In control
You can see everything it did, and undo nothing by accident.
The assistant works through a job on its own, then reports back. Everything it touched is visible, and the decisions that matter stay yours.
It asks before anything permanent
Buying a phone number, deleting an agent, removing a tool connection. You get a card showing exactly what it intends to do.
Approve, correct, or decline
If the details are nearly right, fix them first and it uses your version. If you decline, tell it why and it finds another way instead of guessing.
Every action is logged
A full record of what the assistant did in your workspace, there for you to read whenever you want.
Watch it work
Each step appears as it happens. Open any one of them to see what it actually ran.
Drill into any call
The exact request it sent and the exact response it got back, for any tool call in any run.
See what it spent
Costs broken down by agent, by model, and by type, so there's never a mystery line on your bill.
Where it stops
It works for you, and only when you ask.
- It never acts on its own. There's no schedule and nothing to trigger it. It works while you're talking to it, then stops. For work that should run on a timetable, it will build you a scheduled agent that does.
- It's yours, not your customers'. It isn't something you put on your website, a phone number, or a messaging app. That's what the agents you build are for.
- It's part of every workspace. It can't be deleted or duplicated, it only ever sees your workspace, and it doesn't count towards the agent limit on your plan.