Simple Application Server:OpenClaw application image release notes

March 26, 2026

Released the OpenClaw 2026.3.13 image, which upgrades the built-in software to version 2026.3.13.

This release includes the following updates compared to the OpenClaw 2026.3.3 application image:

• Official plugin upgrade: Upgraded the official plugins for DingTalk, Lark, and WeCom to improve channel stability and functionality.

• WeCom channel optimization: The connection mode is upgraded from URL callback to a persistent connection, improving message response speed and stability.

• Improved configuration experience: Added a validation feature for model configurations to prevent configuration errors.

March 5, 2026

The OpenClaw 2026.3.3 release upgrades the built-in software to version 2026.3.3 and includes the latest community features.

This release includes the following updates compared to the OpenClaw 2026.2.26 application image:

• Built-in headless browser configuration: Pre-configured a managed Chrome headless browser. This allows OpenClaw to use dedicated user profiles for more secure and stable browser operations.

• Pre-configured core enhancement skills: Multiple foundational extension skills are now pre-configured to enhance the AI's intelligence and automation. These include:

  • find-skills: Discovers and installs useful skills.

  • skill-vetter: Performs security scans on skills to ensure a safe runtime environment.

  • agent-browser: Enables browser automation.

  • self-improving-agent: Provides a self-improvement feature enabling the AI assistant to learn from its mistakes.

  • proactive-agent: Enhances OpenClaw's proactivity and initiative.

• Enhanced DingTalk channel capabilities: You can now send attachments, such as images and compressed files (up to 5 MB), to OpenClaw through private messages in DingTalk.

• Lark channel fix: Optimized account configuration for the Lark channel to resolve a reported issue where scheduled tasks failed to run for some users.

February 28, 2026

Released OpenClaw 2026.2.26. The built-in software is upgraded to OpenClaw version 2026.2.26 and includes the latest features from the community.

This release includes the following updates compared to the OpenClaw 2026.2.9 application image:

• Native multimodal model support: Pre-configured the qwen3.5-plus native multimodal model. You can now process both text and image inputs with this single model.

• Memory and stability optimization: Pre-configured 2 GiB of swap virtual memory to prevent out-of-memory errors. (Enabling a swap partition may decrease memory I/O performance. If the instance runs out of memory, we recommend that you first upgrade your plan to increase the physical memory of the instance.)

February 26, 2026 update

• Model and compatibility optimization. Improved compatibility with the latest Coding Plan. On the product page in the Simple Application Server console, you can switch freely between qwen3.5-plus, kimi-k2.5, glm-5, MiniMax-M2.5, and other latest models.

• Added multimodal capability support. Compatible with the image input capabilities of the qwen3.5-plus and kimi-k2.5 models. Now, you can solve both text and image problems using a single model.

February 11, 2026 Release

OpenClaw 2026.2.9 was released and supports the latest community features released before February 9, 2026.

Updates compared to the OpenClaw 2026.2.3 application image:

• Improved web search by prioritizing SearXNG.

• Added built-in plugins for DingTalk, Lark, WeCom, and QQ, in addition to the AppFlow bridge mode. You can configure these services by entering their keys on the application's product page.

• Added a personalization panel for the AI assistant. This panel includes a built-in Markdown editor to directly modify the SOUL.md, IDENTITY.md, and AGENTS.md files.

• You can modify models in the console as follows:

  1. You can switch between different Model Studio models directly from a drop-down menu. This helps users who were unsure how to change models using the OpenClaw configuration file.

  2. You can manually enter specific models supported by Model Studio.

• The initialization process now generates a random service port by default. This effectively hides the endpoint and helps prevent malicious scans.

• Public network access control: A one-click feature was added to enable or disable public network access.

• Gateway restart: A Gateway Restart feature was added. You can restart the gateway service from the user interface.

February 5, 2026 Release

OpenClaw 2026.2.3 was released. It upgrades the built-in software to OpenClaw and supports the latest community features released before February 3, 2026.

What's new compared to OpenClaw (Moltbot) 2026.1.27-beta.1?

• Web search capabilities are now provided by default in all regions, based on the built-in SearXNG web search Skill. No extra configuration or fees are required.

• Support was added for collaborative work between text processing and image recognition. The default Image Model for image recognition is qwen3-vl-plus.

• Docker is now built into OpenClaw 2026.2.3. This meets the requirement for Agents that use a Docker runtime as a sandbox.