From brief to first version.
Turn a software idea into stages, assigned work, and files you can inspect.
Your AI team · One connected workspace
Big ideas need a team.
Meet yours.
Plan, build, and review software, research, and content with an AI team you can direct.
Local-first. Your workspace, on your machine. ↗01 / Your city
Every department, a building.
Every agent, a car.
Each part of your project has a place. Your team drives between them, doing the work.
02 / Plan
Your idea.
A plan your team can follow.
Your project lead turns a rough brief into stages, assignments and review points.
03 / Team · Coding HQ
More than a reply.
A working team.
Open an agent’s conversation, inspect terminal output, and keep files and context close.
04 / Review · Command Citadel
Your team moves.
You make the call.
Answer questions, review results, and decide what reaches your project.
05 / Early access
A world of work.
A team of your own.
Join the list to hear when early access is ready.
Get early access ↗The city · 4 of 24 departments
A place for every part
of your project.
The team
Every agent
has a car.
Each agent drives to the building where its work is. These six are the example project’s team.
01 / 06
Nova
Project lead · Command Citadel
Inside the product · actual app screens
From an idea
to applied work.
- 01Approve the plan
- 02The plan is on its way
- 03Your team is working
- 04Agents report back
- 05They ask when they need you
- 06Review before anything is saved
- 07Applied to your project
Captured from the current local build with an example project. Not a live agent session.
Watch
See it move.
Then step inside.
A 24-second tour recorded in the desktop game: the living city, the Coding HQ team, and Files & review.
Recorded in the game
Your work. A living city.
Keep the context with the result.
Organize source material, drafts, and decisions around a shared goal.
See who is doing what.
Follow assignments, open conversations, and decide what happens next.
Watch an idea become a project
A clear plan.
A team in motion.
Discuss the goal. Review the plan. Follow the work.
Interactive demoYou approve the direction.
Your project lead turns your idea into stages, assignments and review points.
A conversation is just the beginning
Work directly with
every agent.
Open their chat. Follow their work.
Keep the files and context close.
You set the direction
Your team moves.
You stay in control.
Try a decision · example
Local-first by design
Your workspace.
On your machine.
Keep your desktop project files and saved workspace records on your computer. Choose the connections your team uses.
01 / Your space
Keep the work close.
Your projects, saved conversations, and workspace records live with your local desktop setup. Pick the project folder you want to work in.
02 / Your model
Choose your connection.
Use a supported provider or connect a compatible model server, including Ollama, LM Studio, or Open WebUI. Check readiness before starting.
03 / Your boundaries
Decide what connects.
Choose which agents can use a configured service. Review its exact request before sending through the service connection flow.
Local storage. Clear data boundaries.
Connected AI providers can receive your prompts, relevant files, and project context. Approved service requests send data to that service. Using a local model does not automatically make external tools offline.
Desktop is the current setup. VPS builds are planned for the next phase, with the workspace on the server you choose.
Security & control
Your project.
Your boundaries.
Set the pace, choose the permissions, and review what reaches your project.
SECURITY HQ
Clear checkpoints.
In your hands.
Plan approvalTool permissionsChange review
Working copy→Review→Apply
Keep drafts separate.
Agents work in session working copies. Inspect supported changes and send them for review before choosing to apply them to your project.
Decide the boundaries.
Choose when agents can act, set delegation limits, and block individual tools. Available controls depend on the connected provider.
Keep the work visible.
Read recorded activity, inspect tool requests, and return to decisions with their original context. Plan approval and applying files are separate steps.
Recover deliberately.
Review held sessions and interrupted work before continuing. Create project backups and restore to a new folder while keeping the current project.
Know what each protection covers.
Clear scope helps you choose the right setup for your work.
Working copies, commands, and isolation
A working copy separates draft files from your original project. Host commands still run with the worker account’s permissions, which may extend beyond that folder. The optional sandboxed command tool isolates its own checks; it does not place the entire AI provider in a sandbox. Provider support and your saved settings determine which checks apply.
Credentials and connected services
Supported service-integration credentials are saved encrypted and kept out of shared project memory. Storage and setup depend on your platform. Your AI provider’s sign-in and credential handling follow that provider’s own connection setup.
What happens to the context I share?
Connected AI providers can receive your prompts and the project context needed for a request. Their data handling follows their own terms and account settings. Review the files, tools, and services you make available to your team.
Go a little deeper
The details that
keep work connected.
Discover the tools behind the city.
Search a feature, or explore by purpose.
What to expect
Start small.
Stay in the loop.
Early access starts with interest, not an instant download or a promise of a finished project.
- 01
Join the list.
Share your email and preferred desktop platform. No payment is collected by this form.
- 02
Check your setup.
Supported provider connections, system requirements, and pricing will be shared with the release information.
- 03
Bring one clear goal.
Review your team’s plan, set permissions, and inspect the work before applying supported changes.
Windows, macOS, and Linux downloads come later. VPS builds are planned for the next phase. AI provider access and usage are governed by your provider’s account and terms.
Early access
Your place in the city
starts here.
Join the list. We’ll let you know
when early access is ready.
Download · free
- NowEarly-access list
Be among the first to explore Skyline City.
- LaterDesktop downloads
Windows, macOS and Linux downloads will be shared later.
- Next phaseVPS builds
Another way to run Skyline City, planned for the next phase.
A few things to know
Before you
step inside.
What is Skyline City?
Skyline City is a workspace for coordinating AI agents and reviewing their work. Projects, agents and tools have places in a city you can explore, alongside chat, terminals, files and familiar project controls.
What can I work on?
The project flow can organize software, writing, research and other goals into plans and reviewable results. What a team can execute depends on its connected providers, tools, permissions and skills. Building names organize the work; they do not guarantee a specialist capability.
Does the team start work without asking?
The guided flow lets you review the plan before starting the team. Delegation settings, permissions and approval requests help you decide how work proceeds.
What will it cost?
Skyline pricing and early-access terms have not been announced. Joining this list is not a purchase. Connected AI providers may have their own subscriptions or usage charges; check the account and plan you connect.
Can I download it today?
This page is collecting early-access interest. Windows, macOS and Linux downloads will be shared later. VPS builds are planned for the next phase; no release dates are announced here.
Are the examples on this page live agents?
No. The product tour contains actual app captures and a separately labeled internal test result. The interactive walkthrough, conversations and decisions are examples. They show how the experience works without connecting to your projects or running tasks.
Does all my data stay on my computer?
Your local desktop project files and saved workspace records stay with your local setup. Connected AI providers may receive prompts, relevant files, and project context; service requests may send data to external services. You can configure a compatible local model server, but any external tools still have their own data boundaries. Local-first does not mean every workflow is fully offline.
Which AI providers can I use?
The platform includes provider setup and connection controls. Supported configurations and any provider requirements will be detailed with the early-access release. Provider usage costs are separate from the examples on this page.