Try smedja on a real task in your own repository.
We're looking for a small group of developers who already run a coding agent on an API key. You hand smedja one task you'd do anyway, and check what it did.
All five of these are true for you.
- You use a coding agent every day: Claude Code, Codex or Gemini CLI.
- You already have an API key for it: an Anthropic API key for Claude Code, an OpenAI API key for Codex, or an AI Studio API key for Gemini.
- You may let an agent work in your own repository, on GitHub or gitlab.com, or a public repository on any other git host.
- You bring a concrete task, and you can check the result yourself: a diff to read, a test to run.
- You have somewhere for it to run: a Docker Sandboxes account, or your own Linux machine with Docker.
This cohort is for API keys. If you use Claude Code or Codex only through a subscription, you belong to a separate, later cohort. You can still join the waitlist.
Your model usage is billed to your key by your provider. smedja never bills you for tokens. Your key never enters the workspace: model requests go through a proxy, and the workspace holds a placeholder.
Sign in with GitHub or Google from this page. That puts you on the waitlist.
We read each request and choose a small group. Asking doesn't guarantee a place. We email you when your account is ready.
Add your agent with your API key, choose where it runs, connect your repository and describe the task.