Meet your AI coworkers.

Crewlings

Bring Silicon Valley to you.

Your AI crew, together in one office.

Explore the experience

Coming soon

Milo, our orange knitted coworker with headphones

Milo Engineer

Pip, our pink knitted coworker with a green sprout

Pip Researcher

Tess, our ivory knitted coworker with a camera

Tess PR manager

Jules, our purple knitted coworker with sunglasses

Jules Designer

Six coworkers. One shared mission. A working local prototype

The film / 01:00

Your startup.
Your crew.

A familiar place to give AI work.
And see what happens next.

Work has a place now.

Less managing tabs.
More moving things forward.

Name your coworkers. Give them roles. Share an instruction, then follow the work through browser pages, saved files, and teammate handoffs.

01 Your company, your crew

Make the
office yours.

Set your company name, logo, and mission. Create up to six coworkers with names, roles, and the company facts they need to get started.

Crewlings' actual startup office with the NOVA company identity

02 A clear starting point

Ask who's
available.

Free coworkers raise a hand. Choose one and give them a brief. Type an instruction, or record it, review the transcript, and send.

Pip's chat and personal workspace in the actual local prototype

03 Work you can follow

Look over
their shoulder.

Open a coworker's chat to see their browser screen, saved files, and tool activity. Each has a personal browser context and workspace.

An actual browser screen captured while Pip researches Blender

04 A shared next step

Let the work
change hands.

Pip's research can become Tess's announcement. Share findings and completed artifacts within a mission, or queue the next coworker's contribution.

Tess's real workspace showing the saved beta announcement file

Captured from the working local prototype.

Meet your coworkers

Different roles.
One direction.

A name. A role. A place in the team.
You decide who does what.

MiloEngineer

Engineer

Meet Milo.

Give Milo a technical question to research, a specification to work through, or findings to save for the team.

These are starting roles. Names and roles are editable in the app.

An interactive preview

Start with
a conversation.

Pick a coworker. Give them a starting point.
Explore an example of how a task can move through the crew.

A scripted walkthrough. No AI call, recording, or external action.

Behind the little faces

Real work.
A visible process.

The office is the interface.
The work happens underneath.

01 / Context

A shared brief.

Keep company facts and your mission available to the crew. Saved work, messages, and settings persist in the local app.

02 / Tools

Access you choose.

Enable browser and file tools per coworker. Configure Gmail or a supported MCP service, then review actions that need approval.

03 / Runtime

A local starting point.

Use an installed Ollama model. The current prototype runs model requests in order and continues while the local service is running.

Built and tested locally. Gmail and MCP connections require setup. They are not connected in this demo.

A few things to know

Small details.
Clear answers.

Is this a working product or a concept?

Crewlings is a working local prototype. The film shows the actual 3D office, browser screens, task interface, and saved artifacts. Browser work and teammate handoffs have been checked with the installed small Qwen model. Open-ended task quality and long-term autonomy are still being developed.

Can I rename the coworkers?

Yes. Each of the six coworkers has an editable name and role in the app. Those details guide task routing and the model's instructions. You can also set your company name, logo, mission, and company facts.

Do the coworkers work together?

They can share messages and completed artifacts within a mission, and queue another coworker's contribution. Contributions run in order rather than all at once. The current runtime limits a mission to eight contributions.

Does each coworker have its own computer?

Each coworker has a separate browser context, with its own cookies and local storage, plus workspace files. They share one Chromium process. The screen in their chat is a browser capture, not a separate virtual machine or a full remote desktop.

How do voice and connectors work?

Voice uses record, transcribe, review, and send, with a 30-second recording limit. Gmail requires Google OAuth setup. MCP plugins require a supported public Streamable HTTP endpoint and authentication. Gmail and MCP behavior has been tested with fixtures; this demo has no real mailbox or external plugin connected.

Is it always on? Is it fully offline?

Work continues while your local service and model are running. It stops when the machine sleeps or the service closes. Model inference and app storage are local; browser pages, Gmail, plugins, and optional browser speech services can use the network.

Crewlings

Crewlings

Your startup.
Your crew.