Skip to main content
The Kodexa Agent is an AI companion that lives inside the project workspace. You talk to it in a chat pane, and it can read your project’s context, propose changes to your resources, drive the workspace UI, and bind shared resources from your organization into the current project.
This section is about the Kodexa Agent running inside the Kodexa UI. If you author metadata locally in a Git repository, see Working with Claude Code instead. The two surfaces are complementary — the same skills power both.

​
In-UI Agent Or Git-Backed Workflow

Both surfaces let you author Kodexa resources with AI. They differ in where the work happens, what gets touched, and how changes land. You can use both. Many teams iterate in the workspace with the agent, then sync to Git when the project’s shape stabilizes.

​
What The Agent Does

A typical interaction:
  1. You ask the agent something in natural language (“add an invoice_number taxon to the Invoice taxonomy”).
  2. The agent inspects the project, drafts the change, and opens the relevant resource editor with the proposed edit.
  3. You review the diff in the editor, tweak if needed, and click Save to commit. Nothing persists until you save.

​
Pages In This Section

Get started

Find the chat pane, send your first message, and see what the agent already knows.

Configure the companion

Pick a runtime, write a custom prompt, and bind modules in Project Settings.

Channels and context

Task, project, and admin channels — what the agent knows in each.

Manage project resources

Ask the agent to author taxonomies, forms, plans, task templates, and bridges.

Discover and bind

Find shared resources in your org and attach them to the current project.

Work with tasks

Per-task conversations and task-aware context.

UI interactions

Option buttons, document opens, focused attributes, and notifications.

Troubleshooting

Common issues and how to recover.