The AI workspace for Indonesian companies
Projects, team chat, and AI agents. One workspace.
Plan projects, talk with your team, and assign work to AI agents. Keep the conversation, context, and approvals together in KerjaBareng.
Need it on your own servers? Talk to us about self-hosting
- Issues & projects
- Team chat
- AI agents
- Company memory
Actual product screens with example data from Dapur Nusa. Select any screenshot to explore it.
Built for everyday teamwork
- One workspace
- projects, conversations, agents, and company knowledge
- In your browser
- no installation needed to start working
- Visible costs
- track AI usage and set workspace spending limits
- Human approval
- review agent requests before approving the next step
- 2 languages
- Indonesian first, English fully available
The shape of the product
Four parts, one workspace
Not four apps taped together. Issues, chat, agents and company memory sit on the same data — which is why an agent already knows what the team is working on without you re-explaining it.
01 — Work
An issue tracker that stays light
List, board or table; statuses, priorities, labels, staged sub-issues and dependencies. Quick for non-technical people, tidy enough to answer for.
See how it works →02 — Conversation
Team chat, one channel per division
Channels, threads, and direct messages. Mention an agent with @ mid-conversation and it answers in the same thread — no app switching.
See how it works →03 — Workforce
Agents with a role, not a chatbot
Every division gets an agent with its own name, instructions and permissions. It appears in the assignee list next to people, moves the status and reports what it did.
See how they work →04 — Memory
Company context that stays put
Tone of voice, forbidden words, prices and processes stored once. Every agent reads them before it works, so a correction you make today still holds next month.
See an example →The real problem
Why AI adoption stalls in your team
AI used in the shadows
Employees paste company data into personal ChatGPT accounts. No one knows what leaves, and nothing is auditable.
Tools that don't talk to each other
One tool to write, one to summarize, another for tasks — results get copy-pasted by hand, none of it woven into how the team actually works.
Nothing technical on staff
Most AI agent products require installing a CLI, a daemon, and your own model credentials — a non-starter for most teams in Indonesia.
AI amnesia
Every new chat starts from zero. It doesn't know your product, your prices, your customers, or the words your team never uses.
Our promise
AI that shows up already built, already knows your company, and is safe by default.
Not an empty chatbot you have to teach from scratch. Not a daemon you have to install and keep running. KerjaBareng ships agents ready per division, puts them on the same board as your team, and locks every outbound action behind your approval.
The difference
Agents that do the work, not suggest it
The difference between an assistant and a colleague is accountability. Here an agent takes a task on someone’s behalf, is judged by that person’s permissions, and leaves a record of every step.
Assign
Pick an agent as assignee, the way you pick a person
The moment an issue is assigned to an agent, the queue picks it up. The status moves itself to “In progress”, then to “In review” when the result is ready for you.
- The work lands as an ordinary comment on the same issue
- A full transcript: which tools it used, which data it read
- Cancel, re-run or ask for a revision at any point
Put up a fence
Anything that leaves the company waits for a yes
An agent may draft a broadcast, answer a complaint or schedule a post — but it asks first. You approve, ask for a revision, or reject.
- A separate approval queue you can clear with ⌘⏎
- Asking for a revision starts the agent again, with your note
- Every decision is recorded: who, when, and why
Talk to it
Mention an agent in a channel, it answers in the thread
Not every piece of work deserves an issue. For the small ones, mention the agent mid-conversation — the answer sticks to that same thread, not to a separate chat box.
- One channel per division, plus threads and direct messages
- Turn a message into an issue when it turns out to be real work
- Agents never reply to agents — no runaway conversations
An agent only reads the channels it is allowed to read. Private channels and direct messages stay closed unless the agent is explicitly given access.
Getting started
From nothing to a working team, in one afternoon
Nothing to install on anybody’s machine, no API keys for you to buy, no mandatory data migration. You can do the steps below while your coffee cools.
- ± 2 minutes
Sign in and create a workspace
Sign in with Google or a code sent to your work email. Name the company, tick the divisions you want to try, done.
- automatic
Divisions, channels and agents set themselves up
Every division you tick immediately gets a chat channel and one agent with instructions that make sense for that role. All of it is editable later.
- ± 15 seconds
Give an agent one real piece of work
Create an issue, pick an agent as assignee, save. Within seconds the status moves and the answer appears as a comment — with the transcript of how it got there.
- the days after
Correct it once, it holds from then on
Every time you fix an agent’s output, save the rule into Company memory. The next agent reads it first, so the same mistake does not come back.
If you already use another tool, you do not have to switch it off first. Many teams start with a single division here and leave the rest running as usual until they are convinced.
The rest of it
The things that make it usable every day
The features nobody puts in an advert, and everybody looks for in week two.
Three views, one set of data
A list to work in, a board to see the flow, a table to audit. Your choice is remembered per page.
Projects and staged sub-issues
Break big work into stages; finishing one stage can start the agent on the next.
Scheduled routines
Morning recaps, invoice reminders, weekly reports — they run themselves and post into the channel you choose.
Search and ⌘K
One box to jump to a page, find an issue or run a command. Single-letter shortcuts for the things you do often.
Attachments the agent can read
Upload a PDF, image or spreadsheet to an issue; the agent may read it if you allow it.
Costs you can see
AI cost per agent and per run in rupiah, plus a spending cap that stops an agent before it overspends.
The KerjaBareng virtual office
Work from anywhere. Still feel together.
Find teammates on your office map, settle into your own desk, and take a conversation into a meeting. Everything connects to the work already in your workspace.
Team area
Meeting room
Social space
RakaDinaBimaLet's go!MayaTono
SariGreat work!AlyaYou- See who's free
- Wave them over
- Hear nearby conversations
- Join in a click
See who is here
Find teammates by name or room, follow focus cycles, and keep the roster nearby in a compact Mini Mode window.
Spaces that fit your team
Draft a layout, choose a template, and decorate desks. Add room links and status objects for an issue, agent, approval, or signed webhook.
Conversations on your terms
Choose nearby audio, temporary chat, or a meeting. Record with participant consent, replay the private result, and draft AI summaries with source citations.
Sketch ideas together
Open a shared whiteboard from a room, document, or meeting. Add notes, shapes, arrows, and drawings, then export your work.
Guests enter with approval
The host approves every visit to a specific room. A guest link does not open team chat, documents, or saved meeting results.
Learn about guest visitsChoose your focus and sound
Follow shared focus and break cycles. Nearby audio respects distance and room access; your microphone starts only when you choose.
Set up focus and audioTake a break together
React, greet the mascot, or play Together Lights in the same room. Celebration sounds require your permission.
See how to interactFind the conversation you need
Search channel, DM, group, and thread messages with conversation and sender filters. Results follow your current access permissions.
Learn about message search
02 · Skill Packs
Ready-made agents per division, day one
Pick your divisions during onboarding and KerjaBareng wires up the channel and skill pack automatically — not a blank canvas you have to assemble yourself. Your team can delegate routine work immediately, and every finished task feeds back into sharpening that same skill pack.
Kira
Marketing
Caption drafts, content calendar, competitor recaps.
Tania
Sales & CS
Fast replies, lead recaps, follow-up notes.
Rani
Finance & Ops
Daily cash recap, invoice matching, report drafts.
Bimo
HR
Job descriptions, announcement drafts, attendance recap.
03 · Model Hub
The skill picks the model, so you don't have to guess
Every skill already knows which model fits its task best, at the lowest cost. An internal router picks it automatically, your team never needs to learn model jargon, and the cost per task shows up in Rupiah.
| Skill | Model | Est. cost |
|---|---|---|
| Reply to DMs & reviews | Fast model (Terra) | ≈ Rp1,200/reply |
| Write captions & content calendar | Creative model (Fable 5) | ≈ Rp2,800/draft |
| Summarize financial reports | Reasoning model (Opus 5) | ≈ Rp6,500/report |
| Search & summarize internal data | Auto (picked for you) | ≈ Rp900/search |
Cost shows up per task, not just as one total invoice — cost control becomes part of the workflow, not an end-of-month report.
04 · Context layer
Company memory you can see and correct
Every correction you type in chat automatically becomes a new rule for agents. Your team never repeats the same instruction, and anything an agent has learned always shows up as something you can edit or delete at any time.
All of this belongs to your workspace — edit or delete it anytime, and none of it is used to train anyone else's model.
Call customers “Sobat Nusa”
from Dimas's correction
Never use “cheap” — say “great value”
from Bu Rina's correction
Write prices as “Rp25k”, not “Rp 25,000”
from Bu Rina's correction
Updated today
05 · Security & control
Secure by default, not an add-on
Controls that competitors usually reserve for an Enterprise plan ship as standard in KerjaBareng from the Team tier onward.
Approval gate
Sending email/WhatsApp, posting socially, changing financial data — all wait for a click of approval.
Per-agent permissions
Read-only, write to the workspace, or outbound action — you decide for every agent.
Spending caps
A Rupiah ceiling per agent per month; work pauses and waits once it's hit.
Full audit log
Who — or which agent, acting for whom — did what, and when; exportable for PDP-law compliance.
How it works
From sign-up to a finished task
Four steps, no installation required.
Step 1
Sign up
Create a workspace and fill in a short company profile — name, field, products, tone of voice.
Step 2
Pick divisions & agents
Choose the divisions your team uses; their channels and skill-pack agents get installed automatically.
Step 3
Assign or mention
Hand an issue to an agent, or call its name in a channel — it starts working right away.
Step 4
Approve
Review the result in In Review, approve outbound actions in Approval — you make the call.
Why not the others
KerjaBareng vs. what you already use
A short comparison against the three tools most often used instead.
| Aspect | KerjaBareng | Multica | Linear | ChatGPT |
|---|---|---|---|---|
| How agents run | Our servers — nothing to install | A daemon on your machine, needs a CLI | Third-party agents, dev-focused | Chat only, doesn't attach to a task |
| Who it's for | Every division in the company | Engineering / coding-agent teams | Product & technical teams | Individuals |
| Approval before outbound actions | Built in from the Team tier | Not found — just an activity feed | A spend limit, not an action gate | None |
| Team & agent chat | One place — channels, DMs, and agents | Agent chat only, not person-to-person | Separate, lives in Slack | A private session, not for teams |
| Company memory | Visible and correctable by the team | None | None | Gone every new session |
Pricing
Clear pricing from day one
Three plans, one Rupiah-per-seat-per-month unit — no hidden AI fees.
Free
Rp0up to 3 members
- 1 division, 1 agent + Bystep Agent
- 200 issues, unlimited channels & DMs
- 14-day full-feature trial
Team
Rp99,000/seat/month
- Unlimited divisions & projects
- Approval gate, company memory
- Audit log, inbound WhatsApp
Business
Rp199,000/seat/month
- Unlimited agents, premium models
- Per-agent spending caps, SSO
- Audit log & compliance export
Enterprise
Talk to uscustom pricing
- Self-hosting or on-shore Indonesian hosting
- DPA, SLA, priority support
- Custom contract & billing
Launch pricing — will be recalibrated once we measure real AI cost.
Questions
The ones we hear most
Do I need to install anything?
No. Agents run on KerjaBareng's servers with models we already provide — no daemon, CLI, or model credentials for you to set up.
Is my company data used to train AI?
No. Our internal model gateway doesn't store conversation content and doesn't use it to train anyone's model.
What happens if an agent gets something wrong?
An agent's work lands in In Review before it counts as done, and any outbound action — sending, posting, paying — waits for your approval in the Approval queue.
Can an agent send messages or post without my permission?
No. Every outbound action must pass an approval gate — nothing external runs automatically.
Is KerjaBareng ready for Indonesia's PDP law?
Yes. KerjaBareng acts as a data processor, providing a DPA, audit log, and data export — your company stays the data controller.
What if my team already uses Multica, Linear, or a WhatsApp group?
Onboarding takes five minutes with no forced migration — start a fresh workspace and keep WhatsApp as the inbound/outbound channel your team already knows.
What happens if my AI quota runs out mid-month?
Agents stop taking new tasks and you are told; human work is not affected at all. You can move up a plan at any time, and the deferred tasks run once quota is available again.
Is our data used to train AI models?
No. Your workspace content is sent to the model provider only as context while an agent works, and we do not use it for training. The details are on the Security & data page.
Start working alongside your team and AI agents today.
14 days free, no credit card.