FeaturesHow It WorksPricingDocsInstallContact UsGitHub

Parton CLI is in early alpha

Things may break or change between releases. We don't recommend using Parton on production projects yet. If you run into anything unexpected, we'd genuinely appreciate your feedback — it helps us make Parton better for everyone.

Setup & Model Access

Parton must be configured before it can perform AI work.

The setup gate

If you run a command that needs AI and Parton is not configured yet, it blocks execution and starts the setup TUI automatically.

You can also run setup directly:

parton setup

The setup TUI

Setup is a ratatui-based terminal UI with tab-based model selection.

There are four tabs:

  • Default — the fallback model used for any stage that is not explicitly configured.
  • Planning — the model used for the planning stage.
  • Execution — the model used for parallel file generation.
  • Judge — the model used for validation and auto-fix decisions.

Each tab lets you select a provider and model independently.

Supported providers

Parton supports four providers:

  • Claude CLI — uses the claude command locally
  • Codex CLI — uses the codex command locally
  • OpenAI API — direct API access with an API key
  • Ollama — local model inference

Per-stage model selection

Parton can use different models for different stages. This lets you optimize for cost and quality.

Example strategy:

  • Planning — strong reasoning model (e.g. Claude Opus via CLI)
  • Execution — fast coding model (e.g. OpenAI API)
  • Judge — cheap review model (e.g. Ollama local)

You only need to configure Default to get started. Per-stage overrides are optional.

Config file

Setup writes a parton.toml config file.

[models.default]
provider = "claude-cli"
model = "opus"

[models.planning]
provider = "openai-api"
model = "gpt-4o"
env_key = "OPENAI_API_KEY"

[models.execution]
provider = "openai-api"
model = "gpt-4o"
env_key = "OPENAI_API_KEY"

[models.judge]
provider = "ollama"
model = "llama3"

env_key

The env_key field lets you specify a custom environment variable for API keys. This avoids collisions when you use multiple API-based providers or multiple keys for the same provider.

Config behavior

Parton supports:

  • project-local config (parton.toml in the project root)
  • global fallback config

Project config overrides global defaults. This lets you keep a default provider setup globally and override model choices per repo when needed.

Example first run

parton run "create todo app"

If setup is missing, Parton will launch the setup TUI. After configuration:

  1. Provider and model are selected per stage
  2. Config is saved to parton.toml
  3. The run continues into clarification and planning

Common setup issues

Provider CLI not detected

Make sure the CLI tool (e.g. claude, codex) is installed and available in PATH.

Model list does not load

This usually means:

  • provider access is invalid
  • session is expired
  • the provider command failed
  • API key is missing or invalid

API key invalid

Re-enter the key and run setup again:

parton setup

Setup completed but run still blocks

This usually means the config was not saved correctly or the selected model failed validation.

Run setup again to reconfigure:

parton setup

Next