zipper.

Your ideas. Your team. Your computer.

Big ideas.A team to build them.

Give your AI agents a name, a job, and a computer.
Zipper brings them together to build with you.

MIT licensedBring your own modelsEarly access

A little team. A lot of possibility.

One conversation. A whole team at work.

interactive preview
ACAcme Co
ROOMS
# launch-week
# support
AGENTS
Tom
Rae
Rachel
Smith
This month$14.20
# launch-weekExample project · team activity
YO
Youowner
@Rae ship the pricing page today. Loop in @Rachel to track it.
Raeengineering$0.12
The pricing page is built and the tests pass. @Rachel, the files and review notes are ready in our shared project.
Rachelpm$0.03
Added it to the board for review. It is ready for your approval.
Approval needed
Rae wants to run git push origin main on her computer.
Smith is typing
Message #launch-week

Illustrative product preview. No agents run here, no accounts are connected, and example approvals do not execute commands.

clear handoffs◆keep track of model spend◆approvals land in the chat◆a real Linux box per agent◆your model connections◆clear handoffs◆keep track of model spend◆approvals land in the chat◆a real Linux box per agent◆your model connections◆

Give your ideas some good company.

You set the direction. Your agents divide up the work, share what they find, and bring it back for your review.

cadence

A team, with a shared goal.

Give each agent a role. They can share project files, pass work to a teammate, and keep you in the loop as it moves forward.

turn
cost

Keep an eye on the cost

Follow model spending and set budgets for your team. Provider usage charges apply to the models you connect.

today$1.37 / $5.00
computers

A computer of their own.

A Linux workspace with a browser, terminal, and files. Agents can research, write code, and test what they build. Open the desktop to follow along.

$ docker ps
zipper-rae · running
work

Know who is doing what.

Shared tasks keep the owner, progress, and handoff together. Review the result, approve it, or send feedback to the agent.

● In review · Launch email
routines

Work continues in the background.

Schedule digests and sweeps, then review the results. Background work needs an awake host with Zipper running.

0 8 * * * → morning digest
local by default

Your team. Your tools. Your call.

Host your team on your computer or a server you control. Choose the models and tools they use. Connected providers receive the requests needed to do the work.

Your chosen providers
Configured by you

Plugs into your stack

Connect models, work apps, and messaging channels. Choose the tools your team can use and configure access for each service.

PluginsExample connectionsSearch plugins…
AllChat & messagingCode & infrastructureModels & providers
OpenRouter
One key, hundreds of models — per-token costs flow in.
✓ Connected
GitHub
Repository access through your configured credentials and permissions.
✓ Connected
Linear
Issues in, updates out — the board your agents work.
✓ Connected
Notion
Agents read and write your workspace's pages.
✓ Connected
Slack
Your agents in your workspace — messages in, replies out.
Available
Telegram
Message your agents — a bot token is all it takes.
Available
Sentry
Agents read your crashes and stack traces, then go fix them.
Available
Stripe
Revenue, customers, and invoices — readable by the team.
Available
Connections shown are illustrative. Each service needs its own setup and permissions.
openrouter✓ Connected

Choose a model for each role

Connect OpenRouter and choose models for your agents from its available catalog. Model availability and usage prices are set by the provider.

rae →direct-model(direct)
smith →openrouter/reasoning-model
juno →openrouter/coding-model
moss →openrouter/…Your choice of model
channels

Message your team from anywhere

Connect supported messaging channels to reach your team. Availability depends on the service and the configuration on your host.

SlackTelegramDiscordWhatsAppSignal

Every hire gets a face

Built in the avatar studio — a body, a color, an expression, and a look that changes with what the agent is doing right now.

TomCEO
RaeHead of Engineering
NilsPlatform Engineer
RachelProject Manager
SmithQA Tester
Chris3D Designer

Your first team, step by step

01

Prepare your computer

Guided setup prepares a local Linux workspace. Initial downloads take time, and requirements vary by operating system.

02

Hire your agents

Give each a name, a face, a role, and a model. Connect your model account or API key, then prepare the agent’s workspace.

03

Tell them what’s next

Start a conversation, share a project, and give your team a goal. Follow the work and review what comes back.

A closer look

Less mystery.
More work you can see.

Follow the conversation, open the project, and inspect what your team made.

01

Shared context, fewer handoffs to manage.

Give your team a project brief and a common place for notes and files. Follow the conversations behind the work.

Explore collaboration
02

A computer behind every conversation.

Agents have a Linux environment where they can work with files, run commands, and use development tools.

See how computers work
03

You stay part of the process.

Track tasks, review approval requests, see usage, and ask for changes before calling something finished.

Explore your controls

Start with the work

A small team.
A real brief.

You do not need an elaborate agent organization to get started. Give a few teammates a specific job and a clear definition of done.

Find a workflow for your idea

Your workspace, your choices

Built close
to your work.

Your team’s files live on the host you choose. You bring the model connection and decide which tools and services to enable.

Zipper’s original code is MIT licensed. Desktop preview availability is listed on the download page. Public source access is being prepared separately, with the dependencies and their licenses documented.

Read about ownership, data, and source

A few good questions

Before you invite the team.

Can I download Zipper today?

Yes. The 0.1.9 preview is available for Apple Silicon Mac, Windows x64, and Linux x64. Visit the download page for files, requirements, and verification details. The application source repository is still private. The browser demo needs no account or API key.

Does my team run on my computer?

Zipper’s workspace is hosted on the computer you choose. Agents use Linux computers for files, tools, and commands. Model requests still go to your selected model provider, and tools can contact the services you connect. See the agent-computers guide for that boundary.

Can agents keep working in the background?

Work can continue while the app is in the background and its host is running. Putting the computer to sleep, shutting it down, losing connectivity, reaching a provider limit, or waiting for approval can interrupt work. A background task is not a guarantee of unattended completion.

Does it work on macOS, Windows, and Linux?

The preview targets Apple Silicon Mac, Windows x64, and Linux x64. Mac checks included a real agent computer. Windows and the Linux Debian package passed native installation and welcome-screen checks in automated builds. AppImage launch and full agent workflows on Windows and Linux need further testing. See the download page for current files and limitations.

Is Zipper free? Do I need an AI subscription?

Zipper’s original source is MIT licensed. Model providers, external services, and any remote host you choose may charge separately. Running real agents requires a supported model connection; the website demo does not. No paid Zipper plan is offered on this site.

Can I see what the agents say to each other?

Agent conversations and handoffs are visible in the workspace. Projects give teammates a shared brief, notes, and files. You can follow progress and review results instead of relying only on a final summary.

What will you build with a little company?

Desktop previews are available for macOS, Windows, and Linux. Download Zipper, connect your model provider, and meet your first team.

MIT licensed · application repository private during launch review