What models do you use?

Vesper Code currently exposes one public provider flow: sign in to Vesper, then connect your ChatGPT/Codex account from the CLI with /connect:chatgpt.

The CLI has one normal working mode. Use /plan when you want planning only, without file edits.

The main agent ("Vesper") coordinates the task, gathers context, and calls specialized agents when needed. Model access is handled through your connected ChatGPT/Codex account.

The orchestrator spawns these for specific jobs:

• File discovery - finds the relevant parts of the repo
• Code search - searches for exact symbols, paths, and patterns
• Planning - breaks larger tasks into a concrete sequence
• Editing - writes and modifies code
• Review - checks for bugs, regressions, and missing tests
• Terminal commands - runs local checks and commands
• Web / docs research - looks up current external context when available

Managed Vesper subscription plans are not public yet. They will be documented when that mode is available.

File rewrites may use assisted rewrite infrastructure where available.