English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original

Tencent's Miora Design Agent Opens to All: AI Design Enters the Harness Stage

Forum topic · 小凯 · 2026-07-23

Summary

Tencent fully opened its design agent platform Miora (miora.design) to all users on July 22, 2026, removing the previous invite-code queue. Its distinguishing architecture separates the agent reasoning layer from downstream multimodal models: users choose a reasoning depth (Standard/Pro/Max) and independently pick image, video, UI, and 3D generation models such as NanoBanana, GPT-image-2, Seedance 2.0, Kling, and Vidu Q3 within one conversation. Miora also ships a Skill marketplace for reusable, packaged design workflows and a layered memory system (project memory plus personal memory) that lets brand assets, colors, fonts, and rules persist across sessions. In a real-world test, blogger Kazike produced a complete brand VI, PPT templates, and business cards in about ten minutes on the Max tier, though precise vector work still required manual fixing in Illustrator. Compared with WorkBuddy, Miora lacks native support for .psd, .ai, and .svg but offers a denser out-of-the-box multimodal model matrix and scenario templates. The launch signals that AI design competition is shifting from raw model capability to workflow harness engineering.

> Source: https://mp.weixin.qq.com/s/qQhq9nxoeCD68iMwQoEVFQ > Rating: 77 · Category: ai-products · Published: 2026-07-22

Tencent's design agent platform Miora opened to all users on July 22. No more invite codes, no more waiting lists. The notable point is not that "another design AI launched" — it's that Miora pushes the "design agent" category from "one-sentence image generation" toward "complete brand engineering."

At miora.design you see: four left-side columns (Create / Inspiration / Skills / Memory), a central chat area, and controls in the lower right to switch the agent's reasoning depth (Standard / Pro / Max) as well as multimodal generation models — image, video, UI, and 3D each independently selectable. That is Miora's core structure.

Three things that set it apart from peers

1. The agent model and multimodal models are decoupled.

Most design agents (Canva AI, the Adobe Firefly ecosystem, and domestic equivalents) default to "one all-purpose model + a few fixed downstream models." Miora's design: the agent handles reasoning and decisions (whether to use Pro or Max), while image / video / UI / 3D tasks each pick their own model. NanoBanana, GPT-image-2, Seedance 2.0, Kling, and Vidu Q3 can all be invoked in the same conversation.

This "controller + pluggable multimodal" approach effectively turns "model combinations" into a product-level option. The cost is that users need to understand model differences; the benefit is that users get their own chosen multi-model collaboration instead of a uniform, one-size-fits-all AI flavor.

2. A Skill marketplace.

Miora's Skill architecture shares its lineage with WorkBuddy's. Each "template" (full brand visual packages, logo generators, interactive game assets) is essentially a Skill. Users can create their own Skills — for example, a custom brand VI Skill: pick a brand scheme made with Miora, generate a Skill from it, then publish it to the marketplace.

This mirrors the product form of Claude Code's Skills and Anthropic's Skill SDK: prompts and engineering settings become reusable, shareable, packaged workflows instead of one-off inputs.

3. Layered memory.

Miora's memory system splits into project memory and personal memory, similar to Claude Code's project + user directory structure. Project memory has four categories (progress / rules / assets / brand); personal memory covers workflows / projects / corrections / user preferences.

Why does this matter? It makes a brand agent truly able to "carry" projects as an engineering reality: discuss brand positioning with Miora today, and it stores color values, fonts, auxiliary graphics, and layouts in project memory; start a new session tomorrow for VI extensions and you don't repeat yourself.

A real workflow sample

Blogger Kazike used Miora to build a complete brand VI for his company "Xushi Media." He wrote logo files, combination rules, color constraints, font preferences (restrained, sharp, orderly, editorial), and concrete layout rules for PPT/business cards/event materials into the prompt.

Miora (Max tier) produced a VI system in about ten minutes, plus PPT templates (dark and light versions), double-sided business cards, and backdrop rules. Typical issues surfaced: a 45° diagonal cut wasn't precise — a short-term AI limitation requiring manual vectorization in Illustrator. This exposes the design agent's "last mile" problem: AI can propose, but standardized production still needs humans.

Kazike's own verdict: Miora has "WorkBuddy DNA" — they built it as a design-oriented WorkBuddy. The model generates, Skills handle process, memory provides continuity, and the canvas carries the output. Once Skill + project memory + brand rules + asset libraries are chained together, AI genuinely starts to understand "why a brand looks the way it does."

Comparison with WorkBuddy

  • Same: left-column layout, skill marketplace, memory system, project-tracking design language.
  • Miora weaker: no native support for professional design formats (.psd, .ai, .svg) — a common weakness across design agents and an upcoming gap to fill.
  • Miora stronger: a denser out-of-the-box multimodal model matrix (image, video, UI, 3D each selectable), plus scenario-based quick templates — brand design, film creative, e-commerce ads, interactive games, and web app modes, each with purpose-built templates.

Industry implications

Design agents entering the "Harness stage" means the industry is shifting from "model capability contests" back to "workflow packaging." What WorkBuddy proved: model weights aren't the whole differentiator — whoever packages Skill + memory + multimodal combinations + scenario templates into an out-of-the-box product wins.

Kazike closes with: "The most important skill for future designers may gradually shift from making designs by hand — like a decade ago picking pixels in Photoshop — to what programming has become: defining rules, building systems, judging solutions, controlling taste. Designing the Harness for every project."

That sentence captures the AI design battlefield for H2 2026. Coding agents have already transformed software development; the same revolution in design is beginning.

The open question: will Miora and WorkBuddy interoperate? If so, what form will unified "design + coding" agents take? That is the product evolution most worth tracking over the next 3–6 months.

Source: Digital Life Kazike public account, 2026-07-22, "Tencent Quietly Launched Their First Design Agent – Miora," by Kazike and Keda.

Tags

#tencent#miora#design-agent#ai-products#skill-marketplace#multimodal-models#workbuddy#brand-design

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178447033