Skip to main content
This path creates a reviewable manifest, shows the infrastructure plan, applies it, links the current directory, deploys the selected Git ref, and follows the deployment to a terminal state.

​
1. Create an account and activate billing

Sign in at korve.dev. Provisioning requires active organization billing; the CLI shows the published rate card before applying the first plan.

​
2. Install and authorize the CLI

korve login opens a browser authorization page and uses OAuth 2.1 with PKCE. Review the organization, requested role, and exact operation scopes, then approve. See Agent OAuth.

​
3. Initialize from the repository

Run this from a Git checkout whose origin points to github.com/<owner>/<repo>:
Korve detects the repository, current branch, directory-derived project name and slug, and whether the checkout resembles Next.js, Remix, TanStack Start, Vite, Rust, a Dockerfile app, or a generic application. Detection is informational: the generated manifest currently chooses the always-on runtime. Edit and plan the manifest first if you need another runtime class. The command writes korve.toml, prints the complete plan and pricing guidance, and asks:
Confirm only after reviewing the repository, slug, ref, and plan. In CI, provide every value and use --yes only after that review is encoded in the workflow:
If korve.toml or .korve.toml already contains different state, init stops instead of overwriting it. A matching manifest and link are safe to resume. If the matching deployment is already ready, the command prints its URL; if it is still active, the command resumes following it.

​
4. Verify the live application

If the deployment fails, inspect its complete build log and start a durable operator thread:

​
5. Add infrastructure deliberately

Edit korve.toml, then run korve manifest plan -f korve.toml before applying. Add only the databases, storage, queues, crons, caches, alerts, access policy, and maintenance policy the app actually needs. See Infrastructure as code.