Modelway is the platform for coding with AI: one key for every Claude and GPT model in the tools you already use, agents that work on your repositories in the cloud, and a console that accounts for every request.
Works with the harnesses you already have open
Claude Code, Codex, Cursor or your own app. Point it at Modelway once and call any Claude or GPT model by its id.
{
"env": {
"ANTHROPIC_BASE_URL": "https://modelway.vip",
"ANTHROPIC_AUTH_TOKEN": "qk-your-key"
}
}You can sign up with every lab, keep a key per tool and check a dashboard per provider. Or point everything at Modelway once.
Connect a GitHub repository and describe the task. An agent works on it in the cloud, runs the tests and pushes a branch.
Pick a repository and a branch, and say what you want in plain words.
A fresh container clones the repo and runs your setup script. Node, Python, Go, Postgres, Redis and a headless browser are already there.
Each file it reads, each command it runs and each edit, as a diff, while it happens.
It pushes to a branch of its own and never touches your default branch.
See what each model costs you by the hour, keep a monthly budget, and give each tool its own key.
Modelway keeps only a hash of each API key. You see a key once, when you create it.
Upstream keys and sign-ins are encrypted with a key that a copy of the database doesn't include.
Each Code session runs in a container of its own, with no inbound port and no public IP.
Sandboxes reach package registries and GitHub through a proxy that only lets listed domains through.
The agent holds no credential that works outside its session. The real ones stay in the runner.
Agents push only to a branch of their own. Opening the pull request is up to you.
Every plan includes every model. Credits reset on the first of each month, and the console shows where they went.
Every model in the tools you already use.
For people who code with agents every day.
For heavy agent work, all day, in parallel.
Prices in USD. A request that fails before the model answers costs nothing.
Anything else? Write to [email protected].
One account for coding with AI: an API key that works with every Claude and GPT model in the tools you already use, agents that work on your repositories in the cloud, and a console for usage, limits and keys.
Claude Code, Codex, Cursor, OpenCode, Claude Desktop, Cline, Zed, and any app that speaks the Anthropic Messages, OpenAI Chat Completions or OpenAI Responses API. You change the base URL and the key, and nothing else.
Each request costs credits according to the model and the tokens it uses. Your plan's credits reset on the first of every month, in UTC, and the console shows what you spent by hour and by model. A request that fails before the model answers costs nothing.
Requests are refused with a message that names the day your credits reset. You can move to a bigger plan at any time to keep going.
Each session runs in a container of its own with no inbound port, reaches only allowlisted domains, and pushes to a branch of its own. It never touches your default branch: you review the change and open the pull request.
Modelway keeps only a hash of each key and shows you the key once, when you create it. You can have one per tool or machine and revoke any of them from the console.
No. New Claude and GPT models show up in the catalog when they ship and draw from the same credits as the rest.
Sign up with Google. Your key works as soon as your account has API access.