Local and hybrid model swarms stack four layers of complexity on top of your AI workstation. Nightshift absorbs every one of them.
Nightshift stacks five abstractions between you and the work. Models run it. Fleets combine them. Jobs bring it in. Workers read and check it. Methods and skills are what it learns from finishing it.
Download models straight onto your machine, or reach any cloud model with your own API key. Nightshift handles the runtimes, formats, and quirks underneath, so running any mode is hassle free.
Video, image, coding, speech, text. The same job form, whichever model does the work.
A fleet is a set of models that work as one crew. Each model is set to a lane, and the lanes are wired so one model’s output becomes the next one’s input. Nightshift combines different model types to drive any outcome.
Download entire fleets and swap them by use case and need. Want to bulk-create websites? There’s a fleet for that. Want to generate entire shows? There’s a fleet for that. Need 600 files converted from one format to another, then processed? It can do that.
A job is the unit of the night. External apps create them by using Nightshift as a router. You queue custom jobs yourself. Or import the AI conversations you already have, from tools like Claude and ChatGPT, and turn what you’ve been asking for into jobs that finish while you sleep.
While a fleet makes attempts, a crew of workers looks after them. Scribes read each attempt, then summarize and tag it so the whole night is searchable. Judges compare attempts and score which ones are most promising. Verifiers check the finalists against your bar, so only work that holds up reaches your morning.
Workers run on small, fast models alongside the fleet. You set the bar; they do the reading.
All of your jobs can be mined for methods and skills: the steps that recur, the fixes that worked, the sequences that got a result on the first pass. Nightshift extracts them, keeps them, and reaches for them next time.
The more nights it works, the more it knows about your work. It learns and evolves over time, without you writing a prompt for it.
Your apps bring the work. Your agents give the instructions. Your phone puts the controls in your hand.
Swipe to explore · Apps, agents, and your phone →
Connect apps that support a custom OpenAI-compatible endpoint. Point them at Nightshift with your access token, and give your everyday tools a backend for the work that doesn’t need an answer this second.
Any agent that can run shell commands can use the ns CLI. Submit a job, watch its progress, and retrieve the result—without babysitting the cockpit.
Pair your phone by QR code and monitor or steer your workstation over your local network. The same live field, wherever you are on your LAN.