OpenClaw 2026.2.2 has been updated with several key points. Feishu now supports the addition of Feishu/Mark plugins, allowing domestic users to finally connect directly. Contribution from the Chinese community. The Web UI has added an Agent dashboard that allows direct management of Agent files, tools, skills, models, channels, and scheduled tasks on the web page. Previously required command-line operations, now visualized. The workspace memory system has added a QMD backend, allowing agents to remember context at the workspace level. Optional activation. The thinking depth of sub agents can be configured, and default thinking levels can be set globally, or individual agents can be configured separately. If you want to save tokens, you can lower them, but if you need to think deeply, you can raise them. This version has fixed a lot of security issues, including Gateway approval requiring permission verification, Slack command adding access control, skill installation adding SSRF protection, Windows execution whitelist blocking bypass vulnerabilities, and voice call adding incoming whitelist. Safety is clearly being taken seriously. The Chinese document has arrived and we have started translating it into Simplified Chinese. It is still being improved, but at least we have it. Other fixes such as Telegram long polling timeout recovery, Snapchat scrolling position repair, and Agent tool call exception repair are all experience optimization. The most practical features of this version are Feishu support and Web UI dashboard, with a large number of security updates and increased contributions from the Chinese community.