Give your app a teammate that does the work
Agents live inside your apps and act on your real data — answering customers, updating records, booking and following up — under permissions you set, in writing, before they start.
Start building Browse agent templatesAn agent is only as good as what it can reach
GateBuild agents are wired into the app itself — its data, its tools and its channels — not pasted on top as a chat bubble.
Works on your real data
Grant the agent tools to read and write the app's own records — look up an order, update a status, book the slot — each permission switched on deliberately, per agent.
Context it can rely on
Attach the documents that define the job — price lists, policies, product details — so answers come from your material, not guesswork.
Memory across conversations
The agent remembers the customer, the history and where things stand, so people never have to repeat themselves.
Reaches people where they are
Meet customers on WhatsApp, Telegram and LINE, as well as inside the app — on every plan, including Free.
Sets workflows in motion
When an agent finishes a step, it can trigger your automations: send the confirmation, notify the team, schedule the follow-up.
Connected through your accounts
With connectors, the agent acts through your own connected tools — calendar, inbox, CRM — always as you, always revocable.
Brief it like a new hire
Five steps, the same ones a good manager uses: explain the job, set the limits, test, then trust.
-
Describe the job
Write what the agent is responsible for, in the words you would use with a person: the goal, the tone, the edge cases.
-
Choose what it can reach
Pick the data and tools this job needs — a table here, a connector there. Nothing else is visible to it.
-
Set write permissions
For each kind of change, decide: the agent may act on its own, or it drafts the action and a human approves it first.
-
Test on past work
Replay real cases from your history and watch how the agent handles them before any customer sees it.
-
Publish and share
Put the agent live in the app or on a channel, and widen its boundaries as it earns your trust.
Why build the agent here
Your words are the interface
No flowcharts, no code. The brief you write is the configuration — change the words, change the behaviour.
It acts, not just answers
A chatbot tells a customer how to request a refund. A GateBuild agent opens the request, updates the status and books the follow-up — inside the same conversation.
Batteries included
Logins, roles, database, hosting and a custom domain come from the same prompt as the app itself. The agent is born inside a working product, not bolted onto a demo.
Boundaries in writing
Row-level permissions are enforced by the data layer itself: an agent can only ever touch the records its role allows, no matter what it is asked.
Agents people start with
Working starting points — copy one, brief it on your business, and put it to work.
Support agent
Answers product questions from your own documents and opens a ticket when a human is needed.
Booking assistant
Checks availability, books the slot and sends confirmations on chat channels.
Lead follow-up agent
Greets new enquiries, asks the qualifying questions and books the call.
Inventory helper
Watches stock levels, flags what is running low and drafts the reorder.
Powerful, inside a fence
Autonomy without control is a liability. These are the controls every agent runs under.
SSO sign-in
Your team reaches agent settings through your identity provider, not shared passwords.
Roles and permissions
Row-level rules decide what each agent — and each person — can read and change.
Publish controls
Choose who can put an agent live, and to whom: private, your workspace, or the public.
Pre-publish security scan
Every build is scanned before it ships, including the permissions an agent has been granted.
Your data stays yours
What your agents read and write is your business data. It is never used to train models.
Agents are included on every plan
Yes — including Free. Agents draw on the same credit pool as the rest of your workspace; see the pricing page for how the meters work.
Questions, answered
How is an agent different from a chatbot?
A chatbot answers from a script or a document. An agent also does: it reads and writes your app's data, uses your tools and completes multi-step jobs — refund opened, slot booked, status updated — inside one conversation.
How do I stop it doing something unintended?
Three layers: the data layer enforces its permissions no matter what it is asked; sensitive actions can be set to human-approved, so the agent drafts and a person confirms; and you test on private, replayed cases before it ever meets a customer. Start narrow, widen as trust is earned.
What data can my agent reach?
Only what you grant it. Reach is set per agent: this table, not that one; this customer's records, scoped by the same row-level rules as a human role. If a job does not need it, the agent cannot see it.
Which model powers the agent?
An automatic choice by default, matched to the job. On the Builder plan and up you can pick a specific model when a job needs a stronger or faster one.
Are agents really included on the free plan?
Yes. Agents ship on all plans, including Free, within the plan's credit allowances. Model selection is the part that starts at Builder.
Do I need to code to build an agent?
No. The brief, the data it can reach and its permissions are all set in plain language and settings. Developers can go further, but nothing about the standard path requires it.
Where can people talk to my agent?
Inside your app, and on WhatsApp, Telegram and LINE. The same agent, the same permissions, on every channel.
Put an agent to work in your app today
Describe the job, set the boundaries, test it on yesterday's cases.
Start building