Open Source Project

Agent first.
Editor last.

The desktop workspace where AI agents write code in parallel — and you lead the team.

macOS Windows

Why Onward?

Traditional IDEs are editor first — you type, the AI assists. Onward flips this around. Agents do the coding. You plan the work, dispatch prompts, and review what they built.

Onward Agent Workbench — multi-agent parallel workspace
01

Prompt Manager — Your Command Center

Write and arrange prompts before dispatching — plan work like a backlog. Schedule execution, color-tag priorities, and dispatch to agents as they become idle. Manage "what to do" separately from "who does it."

02

Multi-Agent Parallel Execution

Run multiple AI agents side-by-side with configurable grid layouts. Broadcast dispatch to selected agents simultaneously. Each agent gets its own working directory, context, and history. Works with any CLI agent — Claude Code, Codex, or your custom toolchain.

03

Task-Level Full Editor

After agents finish writing code, you review what changed. Each task gets a Monaco-powered editor with Git Diff viewer, Git History, global search, and Markdown preview. One workspace per task — complete visibility into what every agent did.

Review with Precision

The editor is there when you need it — not as the starting point. Git Diff, commit history, and code editing in one integrated view per task.

Onward Agent Workbench — task-level editor with Git Diff

See It in Action

Watch a real workflow: dispatching prompts to multiple agents, monitoring progress, and reviewing the results.

Onward Agent Workbench

Installation

Download the latest release from GitHub for your platform.

github.com/OPPO-PersonalAI/Onward-Agent-Workbench/releases
macOS .dmg, .zip
Windows .exe (NSIS), .zip
macOS: Unsigned App Onward is not code-signed. macOS Gatekeeper will block it on first launch. Go to System Settings → Privacy & Security → click "Open Anyway", or run: xattr -cr "/Applications/Onward 2.app"