AutoClaw V1.9.0 is now available, bringing a stronger model, a more specialized agent experience, and a more complete workflow for product creation and software development.
- AutoClaw V1.9.0 now supports GLM-5.2, with a 1M-token context window that helps the agent understand full project backgrounds, plan multi-step work, and execute complex long-running tasks.
- Auto Design is now available as a dedicated design agent that turns complex product requirements into complete interface design solutions. With one prompt, users can move from product ideas to full UI workflows.
GLM-5.2 in AutoClaw: 1M Context Is Not Just About Reading More Text
This update brings Zhipu AI's latest flagship open-source model, GLM-5.2, into AutoClaw. With support for a practical 1M-token context window, AutoClaw can read more, understand more, and reason across larger bodies of information.
For productivity scenarios, this means users can process dozens of documents in a single task, summarize information, identify issues, and generate structured plans without repeatedly restating the background. For technical scenarios, AutoClaw can better understand relationships across multiple files, helping users analyze projects, modify solutions, and move complex work forward more efficiently.
Consider a complex product requirement document. It may include target users, business flows, page structures, feature boundaries, edge cases, visual requirements, and interaction logic. These elements are tightly connected and should not be split into disconnected fragments.
Product managers care whether the requirement has been fully expressed. Designers care whether pages remain consistent and whether the visual system holds together. Developers care whether interaction logic, states, and edge cases are clear enough to implement. If any part of this shared understanding breaks, teams often face rework later.
When the context window is too short, users have to compress the requirement manually or split the task into multiple rounds. That may keep the model working, but it also introduces common problems: details are lost during compression, page relationships become fragmented, visual consistency drifts across rounds, and users spend more time correcting and aligning the result.
With GLM-5.2's 1M-token context window, AutoClaw can reason about task goals, module relationships, file relationships, and delivery requirements inside a more complete context.
Auto Design: A Full-Workflow Design Agent from One Prompt
Auto Design is one of the most visible new capabilities in this release.
It is designed for a practical use case: turning complex product requirements into design solutions that can be reviewed, refined, and delivered.
Auto Design supports long PRD input. Users can provide a complete product requirement document, and the agent will understand the product structure, page relationships, and functional modules before generating the corresponding interface designs.
For products with multiple modules, users do not need to generate screens one by one. Auto Design can work from the overall product structure and produce a systematic set of pages.
It also supports multi-platform design, including mobile apps, web apps, desktop interfaces, and responsive pages. Based on the same requirement document, Auto Design can help users generate designs that are closer to real multi-platform delivery needs.
After generation, users can continue refining specific parts of the result. For example, a page can be made more visually restrained, a module can be reorganized for clearer information hierarchy, or a specific area can be adjusted for spacing and color. Users do not need to regenerate the entire design from scratch.
The final design can also be imported into Figma while preserving layer structure, allowing design assets to move into a familiar collaboration environment.
Auto Design also includes a design style library, design constraints, and quality checks. It helps avoid common AI design patterns such as excessive decoration, generic purple gradients, and repetitive bento-style layouts. Instead, it applies constraints around typography, color, icons, layout, and overall consistency so the output is closer to a usable design draft.
Real Case Study: From a 12,000+ Word PRD to 100+ Interface Screens
Take the "Tanwei App" as an example.
Tanwei is a food discovery app. Its full PRD contains more than 12,000 words and covers eight functional modules. For designers, this is not a simple task. The product needs to organize content browsing, search, detail pages, favorites, recommendations, user profiles, and other core flows, while also considering state relationships and visual consistency across pages.
In a traditional workflow, a designer would first read the full PRD, break down the information architecture, confirm the page list, build screens step by step, check visual consistency repeatedly, and finally organize everything in Figma for collaboration.
With Auto Design, we provided the full PRD directly.
Powered by GLM-5.2's 1M-token context capability, Auto Design read the complete requirement in one pass and delivered more than 100 interface screens. The generated result was displayed in a grid, making it easy to review the overall page structure quickly.
After generation, users can continue selecting specific regions for local edits. They can also import the result into Figma for further refinement.
Whether you are a product manager, designer, independent developer, or startup team, a fast and reviewable product prototype can significantly improve communication during early-stage product reviews, internal discussions, user interviews, or fundraising presentations. It helps teams move into the next decision cycle faster.
How to Use Auto Design
Download AutoClaw V1.9.0 here: https://autoglm.aminer.cn/autoclaw/updates/autoclaw-1.9.0-cn.dmg
Install AutoClaw, open the app, click Auto Design on the home page, wait for the first initialization to complete, and start creating interface designs from your product requirements.
