Design Agent by Lokuma: a design layer your AI agents can call

What it is. A design intelligence layer that works as a callable tool for AI coding agents, including Claude Code, OpenAI Codex, and OpenClaw. When one of those agents needs to produce a landing page, website, or campaign page, it routes that work through Lokuma, which handles layout decisions, typography, and visual balance rather than leaving the agent to guess.

Why it matters. Coding agents can assemble a working page fast, but the result tends to look like something assembled by a coding agent. The design judgment, what goes where, how hierarchy reads, whether the thing feels considered, is the part that usually falls apart. Lokuma is built by people who made design tools, and it tries to put that reasoning layer back into the pipeline so the output lands closer to something a designer would actually hand off.

How to use it. The workflow sits inside whatever agent setup you already have. You point your agent at Lokuma as a tool it can call during a build task. So if you are using Claude Code to spin up a campaign page, instead of letting it make all the visual calls on its own, it hands those decisions to Lokuma. The agent handles the code and the content logic, Lokuma handles how the thing looks and reads. You review the output the same way you would any agent-generated draft, but you are not starting from a wall of unstyled divs.

Limits. This is a narrow tool with a narrow job: it works when your agents are building pages, and it depends entirely on those agents supporting external tool calls. It is also early, so the range of output types and how much design control you get over specific choices is not fully documented yet. If your work lives outside the agent-driven build pipeline, it does not have much to offer.