Chat Panel

The Integration Kit panel is a chat window powered by whichever AI tool you activated. It looks a lot like Pro’s chat, but it’s the activated CLI doing the thinking, not Nwiro’s backend.

Layout

• Active tool tag at the top shows which integration is driving the chat (Claude Code, Codex, or Local LLM).
• Sidebar holds conversations, separate per tool.
• Chat thread shows the same tool-call cards and permission cards as Pro.
• Input box is where you type. The tool’s own conversation context carries forward.

Sending messages

Just type. The message goes to the active tool, which receives the tool catalog from the Kit — the full 209 tools for Claude Code and Codex, or an auto-tuned subset for Local LLM depending on the model’s tool-calling support (chat-only for models with no tool-calling).

Examples across the tool surface:

Read / find

List every blueprint under /Game/Characters that references DamageComponent.

Edit

In BP_Player, add a Jump input action bound to Space and wire it to a Jump node in the event graph.

Refactor

Rename every material instance prefixed MI_ to MI_Stylized_ and update all references.

Runtime

Start PIE, teleport the player to PlayerStart_Vista, and set the bIntroPlayed blackboard key to true.

Tool cards

Each tool the AI calls shows as an inline card with status, inputs, and output. Click to expand. This is your audit trail. Scroll back to see exactly what was edited when.

Permission cards

Destructive tools (delete_asset, write_file, rename_file, anything that mutates .uasset) show a permission card:

• Allow approves this one call.
• Allow for session approves similar calls until the editor closes.
• Deny blocks this call; the AI usually works around it or reports back.

The permission model matches how Claude Code’s own permission system works, so Claude Code users feel at home.

Conversation history

• Per-tool. Claude Code’s conversations and Codex’s conversations are stored separately.
• Local by default, optional cloud sync. Each machine keeps its own history out of the box. Enable Settings → Cloud Sync if you want chats to follow you across machines.
• Persistent. Conversations survive editor restarts.

What the Kit chat doesn’t have

Compared to Pro:

• No multi-model picker. You’re using whatever model the activated CLI uses (usually the CLI lets you pick between its own model options).
• No Biome Generator. That’s a Pro-only pipeline.
• No cost tracking in the header. Billing happens at the AI provider level; the Kit doesn’t intercept it.
• No Nwiro billing / credits. The Kit is a one-time Fab purchase; AI usage is billed by whichever CLI you activated, not by Nwiro. Your Nwiro account (if you sign in for chat sync) stays free.

Tips

• Restart the activated tool if it stops responding. Some CLIs cache the tool schema at startup; restart via the Activate button.
• Let Claude Code be Claude Code. You’re paying for it; use its strengths. The Kit just extends its reach into Unreal.
• Watch the tool cards. If the AI picks a weird tool, tell it (for example, “use edit_blueprint instead of making a new one”). It’ll adjust.