Agents: Autonomous Multi-Step Workflows in WP Godmode

Until now, WP Godmode worked one conversation at a time. You described what you wanted, Codex built it, you iterated. That loop works well for focused tasks — install a plugin, tweak a template, write a custom shortcode.

But some tasks don’t fit neatly into a single conversation.

“Set up a WooCommerce store with sample products, a branded checkout page, and a contact form.” That’s at least three separate pieces of work, some of which can happen in parallel, all of which need to come together at the end. Doing this manually means juggling multiple chats, copying context between them, and keeping track of what finished and what’s still running.

Agents do this for you.

What agents are

An agent is an autonomous orchestrator. You give it a goal. It breaks the work into sub-chats that run independently — often in parallel. When a sub-chat finishes, the agent gets notified immediately, reviews the result, decides what to do next, and keeps going until everything is done.

You can watch the whole thing happen in real time, or walk away and check back later.

How it works

You start an agent the same way you’d start a chat: describe what you want.

The agent gets to work immediately — reading your site’s current state, creating sub-chats, delegating tasks. If something is ambiguous or it needs a decision from you, it asks. Otherwise, it just runs.

Each sub-chat is a full WP Godmode session — it can read and write WordPress files, query the database, install plugins. The agent assigns each one a specific task and sends it off.

Here’s the key part: sub-chats run autonomously. They don’t pause to ask the user questions. They make decisions based on the instructions the agent gave them. And when they finish, the agent is notified immediately — no polling, no delays.

The agent reads the results, updates its internal memory, and decides what happens next. Maybe it creates follow-up chats. Maybe it’s done and wraps up with a summary. The whole cycle — delegate, collect results, adapt — repeats until the goal is met.

What you see

The agent view has two panels. On the left, the agent’s own conversation: its thinking, the tools it calls, the decisions it makes. On the right, a tabbed panel where you can peek into any sub-chat as it streams, or view the agent’s memory file.

Each sub-chat gets a tab with a status indicator — spinning while active, a checkmark when complete. You can click into any of them to watch the work happen in real time.

When the agent needs something from you — a clarifying question, a decision between two approaches — it pauses and the UI surfaces an input bar. Answer, and it picks up where it left off.

What this changes

The gap between “describe what you want” and “it’s live on your site” gets a lot shorter for complex work. A task that previously required you to run five separate chats, keep context in your head, and manually sequence the work now runs as a single agent with one initial prompt.

Agents don’t replace chats — they orchestrate them. Every sub-chat is still a full WP Godmode session with all the same capabilities. Agents just remove the manual coordination layer.

Agents are available now in WP Godmode.