> ## Documentation Index
> Fetch the complete documentation index at: https://docs.neosantara.xyz/llms.txt
> Use this file to discover all available pages before exploring further.

# Jelma CLI Overview

> Official Neosantara CLI runner to launch and configure AI coding agents locally or on cloud providers instantly.

**Jelma** is the official CLI agent runner from Neosantara designed to launch AI coding agents onto your local machine or into cloud providers (AWS, GCP, Hetzner, DigitalOcean, Daytona, E2B, or Sprite) with a single command. Jelma handles credential injection, endpoint routing, and agent environment configurations automatically.

## Why Use Jelma

| Without Jelma | With Jelma |
| :- | :- |
| Manually edit per-agent settings and JSON files | Run your preferred agent directly in one command |
| Remember distinct base URLs per protocol and plan | Automatically routed based on your API key format |
| Repeatedly configure auth tokens across machines | Credentials are securely stored and injected |
| Repetitive setup on each new machine or VPS | Simply run `jelma <agent> <cloud>` anywhere |

## Supported Coding Agents

| Agent | Description | Protocol |
| :- | :- | :- |
| **Claude Code** | Advanced agentic CLI from Anthropic | Anthropic Messages |
| **Codex CLI** | Terminal coding assistant from OpenAI | OpenAI-compatible |
| **OpenClaw** | Interactive coding TUI with browser automation | OpenAI / MCP |
| **OpenCode** | Open-source terminal coding assistant | OpenAI-compatible |
| **Hermes** | Autonomous coding agent from Nous Research | OpenAI-compatible |
| **Pi** | Lightweight and fast terminal coding agent | OpenAI-compatible |

## Deployment Environments

Jelma supports both local execution and automated cloud provisioning:

* **Local Machine**: Runs directly on your workstation without requiring cloud infrastructure credentials.
* **Cloud Providers**: AWS Lightsail, Google Cloud, Hetzner, DigitalOcean, Daytona, E2B, and Sprite.

## Automated Billing Selection

Jelma detects your active credential type automatically:

* **Pay-As-You-Go (PAYG)**: Using a standard key (`nsk_...`) routes traffic to `/v1` or `/anthropic` with per-token deductions.
* **Coding Plan**: Using a coding key (`nsk_code_...`) routes traffic to `/coding/v1` or `/coding/anthropic` drawing from your prompt quota.

## Model Selection & Routing

By default, Jelma executes against `garda-core` (1M context window). You can override this using the `--model` flag:

```bash Terminal icon="terminal" theme={"theme":{"light":"catppuccin-latte","dark":"ayu-dark"}}
# Using a specific catalog model
jelma claude local --model deepseek-v4.1-flash

# Using a custom Model Route (Coding Plan)
jelma claude local --model my-coder
```

## Quick Start

Install Jelma via your terminal:

```bash Terminal icon="terminal" theme={"theme":{"light":"catppuccin-latte","dark":"ayu-dark"}}
curl -fsSL https://cli.neosantara.xyz/install.sh | bash
```

Launch an interactive prompt or specify agent and target directly:

```bash Terminal icon="terminal" theme={"theme":{"light":"catppuccin-latte","dark":"ayu-dark"}}
# Interactive menu for agent, cloud target, and billing mode
jelma

# Launch Claude Code locally
jelma claude local
```


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.