Give your AI a world to work in.
A desktop home for your AI crew. Agents come to life in a 3D world: give them real tasks, connect their tools, and build workflows across your station.
WHAT IT IS
A living 3D simulation on your machine where every crew member is a real agent doing real work. Recruit a crew, connect real tools and real models, and build the station into a workforce shaped around you — the only limit is what you ask for.
- Placed gear is real capability. A signal dish grants web, a cabinet grants files, a workbench grants a terminal. Agents without their own bay draw on gear anywhere on the station; an agent with its own bay uses the gear in its desk’s room.
- Every bay is another agent, running its own work in parallel.
- The floor is honest. An agent at a lit desk means a live run — the timer on screen is the real one.
You direct from COMMS. They think, browse, build files, and deliver — with your models, your keys, on your machine.
FEATURES
- 01
A crew shaped around you
Recruit specialists, give them personalities, and direct their work from COMMS. Each agent has its own desk and notebook.
Meet your crew → - 02
Workflows you can see
Connect agents with conveyors. Use filters, splitters, and joiners to route real jobs and combine their results.
Build a workflow → - 03
Tools that do real work
Let agents browse, work with files, run code, and use connected services. Placed gear grants capabilities to your crew.
Explore capabilities → - 04
Useful work, on repeat
Keep a successful process as a recipe, schedule a routine, or configure Night Shift for unattended work.
Set up a routine → - 05
Your models. Your machine.
Choose a model provider or use a local model. Your station, conversations, and work history live on your computer.
Choose a provider → - 06
Stay connected to your station
Send tasks through messaging channels, connect external tools, and bring other agent harnesses into your workflow.
Connect your tools →
DOCS & HELP
Get up and running
Downloads are not available yet. See what the first run will look like: connect a model and send your first task.
Start here → THE HANDBOOKFind your next step
Guides and reference organized around what you want to do.
Browse documentation → NEED A HAND?Get unstuck
Find an answer, troubleshoot a problem, or contact support.
Open the help center →FAQ
Is bdibranding free?
The software is free to run. Model usage costs whatever your provider charges — you bring your own API key or run a local model for free (the ChatGPT and Grok sign-ins are in the developer version only), and bdibranding shows you real cost per run. No markup, no subscription required to use your own keys. If you'd rather not manage keys at all, bdibranding Credits are an optional managed plan — never required.
Is the work actually real?
Yes — that's the whole point. Every agent run is a real model call with real tools; files get written, browsers get driven, and the ledger records real cost. The station never renders state the harness can't prove. If it looks like an agent is working, it is.
Do I have to build conveyor belts to use it?
No. Chatting with an agent from COMMS needs no belts at all. Belts are for real jobs — work with several parts, or work that should flow through several agents on its own, especially while you're away. The conveyor guide shows the whole idea in five minutes, and the LINES tool drops a finished line in one click.
Where does my data live?
On your machine. Station state, transcripts, memory, and ledgers live in your local bdibranding workspace. Your prompts go to your model provider when an agent runs. On a fresh station an agent can also search the web and read web pages on its own, without asking: the search engine sees the query, and the site sees the request for its page. What asks you first is a command it wants to run on your computer and a web request it sends itself (to an API, for example); connectors and chat channels are contacted only once you add them. At startup it also downloads two public model and price lists that carry nothing of yours. The privacy page lists every outside contact.
What do I need to start?
A Windows or macOS machine and one model provider: an OpenRouter API key works everywhere, and Ollama runs free and fully local. That's it — full power from minute one, nothing gated behind progression.
Can it spend money without me noticing?
No. Every conveyor line has a per-job budget ($2.00 unless you raise it, $50 ceiling) and an optional per-day cap; SETTINGS › SPENDING LIMITS adds hard per-run, per-agent, per-day, and global caps the sidecar enforces against the real ledger. Night Shift is limited to a number of unattended jobs per day you choose. Use Stop in COMMS to interrupt the current task, or type /halt in COMMS to stop every agent, routine, loop and the night shift at once.