Use cases
A lane doesn’t know what it’s running. If it speaks the OpenAI or Anthropic API, from an agent farm to a support bot to a script that reads the web all night, it runs on the same flat plan. Each example below is sized in lanes, so you can see which plan it fits.
Pre-launch. Accounts are open; the API and checkout open soon.
Coding
Explain this function, fix that bug, write the test you were going to skip. An agent in your editor when you feel like it, with nothing ticking in the background.
Fits Solo3 of 4 lanes · $18 / month
Dependency upgrades, flaky tests, migrations, docs nobody wrote. Queue them up at the end of the day and review the branches in the morning.
Fits Pro12 of 16 lanes · $54 / month
An orchestrator hands out tasks, dozens of agents work in their own worktrees, and whatever passes the tests gets merged. The kind of workload per-token bills punish most.
Fits Scale60 of 64 lanes · $180 / month
In your product
Answers from your docs and the customer’s account, drafts replies for your team, and hands off to a person when it should. Busy days cost the same as quiet ones.
Fits Pro10 of 16 lanes · $54 / month
The calls your app makes for its users: summarise this thread, answer from our help centre, tag this upload, draft that email. Ship the feature without pricing it per token.
Fits Pro + 8 extra lanes24 of 24 lanes · $102 / month
Automation
New papers, price changes, job posts, competitors’ changelogs, mentions of your product. Your script fetches the pages; the model reads them and tells you what changed.
Fits Solo4 of 4 lanes · $18 / month
Pull fields out of PDFs and emails, classify tickets, dedupe a CRM, label a backlog. Point the queue at your lanes and let it run until it’s done.
Fits Solo + 4 extra lanes8 of 8 lanes · $42 / month
Just for you
A chat UI on your own server, like Open WebUI or LibreChat, connected to your notes, calendar and home lab. Ask it anything, as often as you like.
Fits Solo4 of 4 lanes · $18 / month
Game NPCs, lesson plans, a bot for your Discord, translating your docs. If it calls a model, count how many requests run at once and which models they use. That’s your lanes.
Examples are illustrative loads at normal speed and context, not measurements. A lane map shows one letter per lane for the model holding it: G GLM 5.3, F GLM 5.3 Flash, D DeepSeek V4.1 Flash, K Kimi K3; dots are spare.
Pick a plan by how much runs at once. What runs through the lanes is never billed per token.