What's New

Updated on 2026-11-11 GMT+08:00

The tables below describe the functions released in each CodeArts Agent version and corresponding documentation updates. New features will be successively launched in each region.

November 2026

November 10, 2026

No.

Feature

Description

Phase

Document

1

Help documentation addition: progressive knowledge

Added the progressive knowledge page of CodeArts Agent.

Commercial use

Progressive Knowledge

October 2026

October 29, 2026

No.

Feature

Description

Phase

Document

1

1M context for GLM-5.2

Expanded the GLM-5.2 context window from 200K to 1M tokens, so an entire repository can be loaded at once, long-running tasks run without interruption, and less information is lost during compression. Added context usage visualization. 

Commercial use

Adding Context > Context Compression

2

Support for the Goal capability

Added the goal command, facilitating nonstop development. 

Commercial use

/goal Command

3

Mermaid image function enhancement and HTML preview

  •  Enhanced Mermaid charts: added export as SVG or PNG, and supported full-screen viewing and zooming.
  • Optimized HTML code blocks: supported live preview rendering, a switch between source code and preview, and full-screen viewing.

Commercial use

How Do I Save an Image Generated by Mermaid to My Local PC?

4

Skill auto-iteration

Skill auto-iteration is an AI capability for continuous self-improvement. When enabled, after completing each task for you, the AI reviews how well it performed, retaining effective methods and refining areas that fell short. The more it is used, the more familiar it becomes with your habits, making it increasingly efficient at handling similar tasks over time. This saves you from manually updating settings. 

Commercial use

User Guide (IDE) > Slash Commands

User Guide (IDE) > Settings > Global Settings of the IDE > Lab Features

User Guide (CLI) > TUI > Slash Commands

User Guide (Extensions) > Settings > Basic Settings > Lab Features

5

Feedback entry added to CodeArts Agent CLI

Added the /report-center command in CodeArts Agent CLI, allowing you to report non-compliant content and provide feedback on issues. 

Commercial use

Slash Commands

6

User experience and interaction optimization

  • Added support for mainstream programming languages in the unit test agent, including Python, C, C++, Go, TypeScript, and JavaScript.
  • Introduced tool self-check and a three-tier dynamic loading mechanism (persistent/deferred/retrieval-based) upon agent startup, reducing incorrect tool calls and conserving context.
  • Enhanced the agent's proactive documentation retrieval capability, automatically invoking the documentation skill to return operational guidance.
  • Added support for reading image content to assist UI design, improving the console's image recognition experience.
  • Multitasking in SDD mode: supported multitask running across SDD mode and normal mode. 
  • Chat export page: moved the select all checkbox above Export Selected, and added tooltips that appear on hover.
  • Fixed an issue where the prompt was inaccurate when CPU usage was too high.
  • Fixed an issue where the AI chat box failed to load properly in specific scenarios.
  • Fixed an issue where a benefits model reported an error in the intelligent Q&A mode.
  • Fixed an issue where words wrapped abnormally.
  • Fixed an issue where a prompt covered a button, making it unclickable.
  • Fixed an issue where federated login did not remember the username.
  • Fixed an issue where users were automatically logged out because of IP address drift.
  • Fixed an issue where tasks ran abnormally because of kernel process disconnection.

Commercial use

User Guide (IDE) > Agents > Agent Chats > Unit Test Agent

Unit Test Agent

Chat History

Multi-Chat Management

7

Security capability enhancement

  • File deletion tool: added directory-level deletion and recovery (using a recycle bin mechanism), added support for Windows, macOS, and Linux, and added a recommendation to use self-developed commands based on security checks. 
  • Improved the experience for high-risk commands: added a second AI risk assessment, allowed or blocked each command based on the result, and displayed why the command is risky on the front end.

Commercial use

8

Documentation improvements

  • Added "How Do I Adjust the Model's Thinking Effort?" to FAQs, describing how to adjust a model's thinking effort. 
  • Updated the "Context > Adding Context" section in User Guide (IDE), and optimized the content in "Context Compression."
  • User Guide (IDE): optimized the "Browser" section.
  • User Guide (IDE): optimized the "Models" section and added a table comparing the differences between adding custom models on the console and the client.
  • Best practice: developing a travel assistant using the CodeArts toolchain plug-in.

Commercial use

How Do I Adjust the Model's Thinking Effort?

Adding Context > Context Compression

Browser

Developing a Travel Assistant Using the CodeArts Toolchain Plug-in

October 10, 2026

No.

Feature

Description

Phase

Document

1

Document optimization

  • Optimized the "What Is Huawei Cloud CodeArts Agent" section in Service Overview and added descriptions about product forms.
  • Updated the "Certificates" section in Service Overview.

Commercial use

Service Overview

September 2026

September 18, 2026

No.

Feature

Description

Phase

Document

1

Project-level cross-chat memories

Added the ability to retain identity preferences, project milestones, collaboration feedback, and external reference notes. New chats can pick up core project context automatically, without needing to repeat information.

Commercial use

Memory

2

Enterprise knowledge management with automatic experience capture

  • Added automatic capture of individual experiences from day-to-day coding, bug fixes, and code reviews to build a dedicated knowledge base.

  • Introduced automatic identification of frequent bug patterns, key technical decisions, and best practices within a project to build a project-level knowledge base.

  • Introduced a daily automated reflection process that surfaces and records new experiences, periodically generating experience summaries and reusable reports.

Commercial use

Experiences

3

Chat token details

Added real-time display of input tokens, output tokens, cache-hit tokens, and cache-hit ratio for a chat.

Commercial use

Adding Context

4

Multimodal input: removed image upload limit

Upgraded multimodal input capabilities and removed the image upload count limit, enabling multiple images to be uploaded at once.

Commercial use

Multimodal Input

5

Codebase retrieval optimization

  • Optimized the vector algorithm, significantly improving code semantic search accuracy.

  • Overcame scale limitations to enable efficient retrieval across large codebases.

  • Built a global code dependency graph, supporting full-chain call tracing and in-depth module architecture analysis.

  • Introduced a new hybrid "vector + graph" dual-engine retrieval system, delivering more precise code search results.

  • Reduced token consumption while optimizing both retrieval efficiency and operating cost.

Commercial use

Repository Index

6

Model visibility scope and team configuration

Added the ability to customize visibility scope and visible teams when creating a new model.

Commercial use

Configuring Custom Models

7

Improved federated account login experience

Addressed the cumbersome federated login process by adding a client-side login window that remembers basic account information, eliminating the need to re-enter key details such as the tenant name.

Commercial use

Quick Start

8

Console experience upgrade

Unified the interaction style across bulk operations and enable status labels.

Commercial use

Knowledge Space Management

9

Context usage display

Redesigned the context usage panel with a card-based visual layout, presenting token consumption by dimension in a more structured way.

Commercial use

Adding Context

10

UI update

Refreshed and optimized the light theme for CodeArts Agent IDE.

Commercial use

11

Streamlined free package activation

Simplified activation: Users can now start using CodeArts Agent simply by logging in with authorization. Users without an active package are automatically activated on the free package upon login, eliminating the need to place an order.

Commercial use

IDE

12

Decoupled enterprise administrator role from Tenant Administrator permissions

Decoupled the enterprise administrator role from IAM's Tenant Administrator permissions. IAM users with Tenant Administrator permissions are no longer automatically granted the enterprise administrator role by default; only the account that purchased a package is granted the enterprise administrator role by default.

Commercial use

Members

13

Improved enterprise custom model configuration

Improved enterprise custom model configuration. Models take effect on the client in real time and are ready to use immediately after setup.

Commercial use

Configuring Custom Models

14

Improved scheduled task experience

Scheduled tasks are automatically enabled upon creation, reducing manual startup steps and improving efficiency.

Commercial use

User Guide (IDE) > Scheduled Tasks

User Guide (CLI) > Scheduled Tasks

15

CodebaseWiki

Added embedded Wiki documentation with one-click generation, incremental updates, and code synchronization, facilitating both human reading and agent memories.

Commercial use

CodebaseWiki

16

Visual Studio Code/JetBrains extension support for macOS

The extensions for Visual Studio Code/JetBrains support macOS and are compatible with both ARM64 (Apple Silicon) and x86_64 (Intel) architectures.

Commercial use

Visual Studio Code > Constraints

JetBrains > Constraints

17

Improved file change display style

Improved the file change dialog, with the accept button highlighted for better visual clarity.

Commercial use

Core Development Modes

18

Improved tool call display

Simplified the presentation of tool call processes, reducing visual clutter and improving interface clarity.

Commercial use

Diff View

19

Improved chat export

Historical chats can be exported including the user's input prompts.

Commercial use

Chat History

Multi-chat Parallelism

20

URL toggle for custom model configuration

Added a toggle for whether to use the full URL address when configuring a custom model, letting you enable or disable it as needed for more flexible configuration.

Commercial use

Adding Custom Models

21

Experience and detail optimization

  • When adding a file/folder as context, you can click it to jump directly to its containing directory.
  • When creating a custom agent, you can freely specify the basic model.
  • You can cancel an in-progress prompt optimization task at any time.
  • Added the ability to adjust model reasoning intensity, letting you balance AI reasoning depth against response speed as needed.
  • Improved the prompt shown when context is too long and gets compressed.
  • Windows: Fixed an intermittent issue where the login page did not automatically close after a successful login.
  • Windows: The model list refreshes automatically after adding a custom model, no longer requiring a manual click on the refresh icon.
  • Windows: After a custom agent is deleted, the system automatically switches to the primary agent, improving workflow continuity.
  • Fixed an issue where model data loaded too slowly in certain scenarios.
  • Fixed an intermittent issue where the model list failed to load.
  • Fixed an issue where individual edition users could not log in to the console.
  • Fixed a request parameter error related to chat_id in the Q&A process.
  • Optimized some functions and fixed some known issues.

Commercial use

Supported Context Types

Creating a Cloud Custom Agent

Intelligent Q&A > Input Optimization

Agent Chats > Input Optimization

22

Documentation improvements

  • Getting Started: improved the document structure
  • User Guide (Space): consolidated all its content into the "Space Mode" chapter of User Guide (IDE).
  • User Guide (IDE): updated the "Memories" chapter by improving the "Constraints" section.
  • User Guide (IDE): updated the "Adding Context" chapter by improving procedures.
  • User Guide (IDE): optimized "Constraints" in the "Models" section.
  • User Guide (Console): optimized "Constraints" in the "Configuring Custom Models" section.
  • Best Practices: optimized the document structure in the "Developing a Crawler Program with CodeArts Agent and a Third-Party R&D Toolchain" section.

Commercial use

Getting Started

Space Mode Overview

Memory

Adding Context

September 01, 2026

No.

Feature

Description

Phase

Document

1

Official launch of the free edition

Officially launched the CodeArts Agent free edition, allowing users to experience its core capabilities for free.

Commercial use

Service Overview

Billing

Getting Started

User Guide (IDE)

August 2026

August 26, 2026

No.

Feature

Description

Phase

Document

1

Commercial launch on Huawei Cloud (international)

Officially launched CodeArts Agent for initial commercial use on Huawei Cloud (international).

Commercial use

August 05, 2026

No.

Feature

Description

Phase

Document

1

Document optimization

  • User Guide (IDE): optimized "Constraints" in the "Models" section.
  • User Guide (Console): optimized "Constraints" in the "Configuring Custom Models" section.
  • Best Practices: optimized the document structure in the "Developing a Crawler Program with CodeArts Agent and a Third-Party R&D Toolchain" section.

OBT

Models

Configuring Custom Models

Developing a Crawler Program with CodeArts Agent and a Third-Party R&D Toolchain

2

Release of 26.7.1

Optimized some functions and fixed some known issues.

OBT

July 2026

July 30, 2026

No.

Feature

Description

Phase

Document

1

CodeArts Agent launched in public beta on the Huawei Cloud international website

CodeArts Agent is an AI-powered coding agent launched by Huawei Cloud, designed for both individual developers and enterprise teams.

  • Deeply integrates IDE and specification-driven autonomous development modes, powered by the flagship GLM-5.2 code model with enhanced long-context understanding and engineering-grade coding capability.
  • Supports building local personal indexes and cloud shared indexes for deep repository understanding.
  • Supports multi-agent collaboration in the agent team mode, offering core capabilities like project-level code generation, R&D Q&A, and unit test generation.
  • Leverages MCP open extensibility to integrate with third-party DevOps toolchains, efficiently boosting R&D productivity.

OBT

Service Overview

Billing

Getting Started

User Guide (IDE)

2

Service Overview

Introduces core capabilities, features, and typical scenarios of the product for a quick product overview.

OBT

Service Overview

3

Billing

Explains pricing rules and helps resolve billing-related queries.

OBT

Billing

4

Getting Started

Guides you step-by-step from zero to hands-on usage via basic tutorials.

OBT

Getting Started

5

User Guide (IDE)

Covers IDE configuration and development workflows for local setup and efficient integration.

OBT

User Guide (IDE)

6

User Guide (Console)

Outlines console modules and workflows to efficiently manage project resources and permissions.

OBT

User Guide (Console)

7

User Guide (Space)

Introduces the key features and typical scenarios of CodeArts Agent Space for smart coding and R&D tasks.

OBT

User Guide (Space)

8

User Guide (CLI)

Explains visual development via Terminal User Interface (TUI) and high-efficiency command execution via Command Line Interface (CLI).

OBT

User Guide (CLI)

9

User Guide (Extensions)

Describes the installation process and introduces the smart coding, retrieval, and extensibility capabilities of CodeArts Agent extensions for Visual Studio Code and JetBrains series, facilitating R&D in your local editor.

OBT

User Guide (Extensions)

10

Best Practices

Offers reusable guides based on typical scenarios, practical tips, and deep dives into advanced capabilities.

OBT

Best Practices

11

FAQs

Gathers high-frequency issues and solutions for quick troubleshooting.

OBT

FAQs