Software that doesn't wait to be clicked.
Clove Studio builds AI agents for growing businesses in Indonesia. An agent reads a request, decides what needs to happen, works across your tools to do it, and checks the result. Your team steps in only when a decision needs a person.
Example run, shown for illustration.
A chatbot answers. An agent gets it done.
Most businesses already have tools for chat, spreadsheets, email and social media. What they don't have is something that moves work between them. People end up copying, pasting, checking and replying all day. That's the gap an agent fills.
- Replies with pre-written answers
- Stops at "please contact our admin"
- Can't see your stock, orders or calendar
- Forgets the conversation once it ends
- You still do the follow-up work
- Understands requests written the way customers really write
- Checks your real data before answering
- Takes action: updates sheets, sends emails, drafts posts
- Remembers context about customers and past work
- Hands off to a person when a decision needs one
Every agent runs the same loop.
Underneath every agent we build is a cycle that repeats until the task is finished. It's what lets an agent handle work that isn't the same twice.
Read
Takes in the request: a message, an email, a file, or a scheduled trigger like "every Monday at 8".
Plan
Works out the steps needed and which tools to use, based on your rules and what it knows about your business.
Act
Carries out each step in your real tools: looking up data, writing replies, updating records.
Check
Reviews the result against what was asked. If something is off or unclear, it tries again or asks a person.
Then it loops back. An agent handling an order might go around this cycle four or five times: read the message, check stock, reply, log the order, confirm.
Agents we build
Each agent is designed around one real job in your business. Start with one, and add more once it's proven.
Customer chat agent
Handles conversations on WhatsApp, Instagram DMs and your website at any hour.
- Product, price and stock questions
- Takes orders and sends payment details
- Follows up on unpaid orders
Operations agent
Takes over the admin that quietly eats your team's week.
- Keeps spreadsheets and records up to date
- Matches payments to orders
- Flags anything that doesn't add up
Content agent
Prepares social content in your brand's voice, ready for you to approve.
- Captions, hooks and post ideas
- Short video scripts
- Replies to comments
Inbox agent
Reads incoming email, sorts what matters, and drafts replies for you to send.
- Sorts by urgency and topic
- Drafts replies in your tone
- Pulls attachments into the right folder
Reporting agent
Turns scattered numbers into a short summary you'll actually read.
- Daily or weekly sales summaries
- Spots best and worst sellers
- Sends it to you on schedule
Custom agent
Have a workflow that doesn't fit a box? We'll design an agent around it.
- Mapped from how you work today
- Connected to the tools you use
- Built for your rules and limits
What makes our agents reliable
An agent is only useful if you can trust it with real work. These are the parts we build into every one.
Connected to your tools
Agents work inside the apps you already use, such as WhatsApp, Instagram, Google Sheets, Gmail and Notion, so you don't have to move to a new system.
Memory of your business
Your products, prices, policies and tone of voice are written into the agent's knowledge, so its answers match how you'd answer.
Clear rules and limits
You decide what an agent may do on its own and what needs approval. Refunds, discounts or anything unusual can always go to a person first.
Bilingual by default
Agents read and write natural Indonesian and English, including the casual, abbreviated way customers actually type.
Every action logged
Each step an agent takes is recorded, so you can see exactly what it did, when, and why.
Handoff to people
When an agent isn't confident, it stops and passes the task to your team with a short summary, instead of guessing.
Where agents fit first
Agents pay off fastest where the same kinds of requests arrive every day and the answers live in data you already have.
Online shops
High chat volume, stock questions, order tracking.
Clinics and pharmacies
Appointment requests, opening hours, general product information.
Creators and agencies
Brand inquiries, content planning, DM requests.
Food and beverage
Orders, reservations, menu and delivery questions.
What's inside an agent
Every Clove Studio agent is built in layers. Each one can be adjusted on its own as your business changes.
Channels
Where requests come in: WhatsApp, Instagram, email, your website, or a schedule.
Instructions
Your rules, tone of voice, and what the agent may and may not do.
Reasoning model
Claude by Anthropic reads the request, plans the steps and writes the responses.
Knowledge
Your products, prices, FAQs and past conversations, kept up to date.
Tools
Connections that let the agent act: reading sheets, sending messages, creating drafts.
Checks and logs
Review steps, approval gates and a full record of everything the agent did.
From first call to a live agent
We start with one workflow and a small test, so you see results before committing to more.
Map the work
We look at how the task is done today, where time goes, and where mistakes happen.
Design the agent
We agree on what it handles alone, what it hands to a person, and which tools it connects to.
Build and test
We build it and test it on real past examples from your business before it talks to anyone.
Launch and improve
It goes live with your team watching. We review its work and keep tuning it.
Built to be trusted with real customers.
An agent speaks for your business. We design every one so you stay in control of what it says and does.
- You approve the risky thingsPayments, refunds, discounts and anything unusual can require a person's approval before the agent acts.
- It says when it doesn't knowAgents are instructed not to invent answers. Unclear requests go to your team.
- Your data stays yoursAgents only access the tools and data you connect, and only what each task needs.
- Nothing happens in the darkEvery message and action is logged, so you can review any conversation at any time.
- Health questions go to professionalsFor clinics and pharmacies, agents handle admin and general information only, never diagnosis or dosing advice.
Why we build on Claude
The model at the centre of an agent decides how well it understands, plans and follows your rules. We chose Claude by Anthropic for our agents.
- Follows detailed instructions and business rules closely
- Handles multi-step tasks and tool use well
- Writes natural Indonesian and English
- Built by a company focused on AI safety, which matters when an agent talks to your customers
Questions
Will an agent replace my staff?
No. Agents take over repetitive tasks so your team has time for the work that needs people: handling special cases, building relationships, and making decisions.
What happens if the agent makes a mistake?
Every action is logged and can be reviewed. Risky actions can require approval first, and you can pause an agent at any time. We use mistakes to improve its instructions.
Do I need to change the tools I use?
Usually not. We connect agents to the tools you already use. If a tool can't be connected, we'll tell you before we start.
How long does it take to build one?
A first agent for a single workflow typically takes a few weeks from first call to launch, depending on how many tools it needs to connect to.
Can it speak like my brand?
Yes. We write your tone of voice into the agent's instructions using your existing messages and posts as reference, and adjust it with you during testing.
What should your first agent do?
Tell us about one task your team repeats every day and how it's done now. We'll reply with whether an agent is a good fit and what it would look like.
team@cloveestudioo.onlineEmail us your workflow