Quick Start

If you only want the shortest path to a working setup, start with Get Working Fast. This quick start is broader — it assumes you want both a first success and a light tour of the interface.

1. Install & authenticate

curl -fsSL https://omegon.styrene.io/install.sh | sh
omegon auth login openai-codex
om

Need nightly builds? Use --channel=nightly with the install script. See Installation for all options.

om starts in slim mode: you get the core coding loop — prompt, edit, validate — with project memory working in the background. Design tree, OpenSpec, and parallel execution are there when you need them, but they stay out of the way until you ask.

Launch omegon instead if you want the full harness surfaces immediately. You can always switch live with /warp, or peek at everything with /help all.

2. Your first prompt

Type a prompt and press Enter:

Read README.md and summarize the project structure

The agent will call the read tool, show the file contents in a tool card, and respond with a summary.

3. Read the footer honestly

Row	What It Means
provider	Concrete runtime provider identity, such as `OpenAI/Codex` or `Anthropic/Claude`
model	Selected runtime model, such as `gpt-5.4`
limit	Provider quota telemetry or bucket metadata when the upstream exposes it
state	Context class, fill percentage, and window size
session	Turns and cumulative session token counts

That distinction matters. The limit row is provider telemetry, not a second model selector.

4. Key commands

/model         # Switch model (opens selector)
/think high    # Raise reasoning effort
/context clan  # Raise context class
/status        # Check provider and system status
/warp          # Toggle om ↔ omegon live
/tutorial      # Launch the interactive overlay tutorial
/update        # Inspect and install a newer release

5. Project memory

Omegon persists durable project facts across sessions. In a fresh repo, start with:

/init

The agent will orient itself to the project, store foundational facts, and carry that memory into future sessions.

6. Multi-file changes

When a task spans multiple files, Omegon can decompose it into parallel branches — each running in an isolated git worktree. This is cleave, and it's available in full mode (omegon) or after /warp from slim mode.

Refactor the auth module to support OAuth2 PKCE flow

# If warranted, Omegon will recommend /cleave.
# You can also trigger it explicitly:
/cleave

If you started with om, cleave isn't shown in /help — that's intentional. Type /cleave directly or /warp to switch to full mode when you're ready for orchestration.

7. Browser surfaces

/auspex open

Launches Omegon's primary browser surface. If the local compatibility bridge is not running yet, Omegon starts it first. For direct local compatibility/debug access, /dash remains available.

8. Tutorial

/tutorial is not a fake tour. It is an interactive overlay that can read code, store memory, and create lifecycle artifacts in your real project.

/tutorial        # current-project path
/tutorial demo   # scripted showcase path

Next steps

Explore at your own pace. Each step builds on the last:

Interactive Tutorial — guided walkthrough of the coding loop, memory, and more
Ecosystem & Integrations — MCP servers, plugins, IDE rules, credentials from other tools
OpenAPI Tools — REST API tools from project OpenAPI specs
Command Reference — every slash command (/help all in-session)
Providers — auth modes, boundaries, and runtime identity
Design Tree — structured architecture exploration (full mode)
Cleave Orchestrator — parallel multi-file execution (full mode)
OpenSpec — spec-driven development lifecycle (full mode)
Sentry Automation — long-running triggered task execution
FAQ — common questions, direct answers