Changelog

This release gives every incident an AI coordinator that decides what to do next and coordinates investigations. As new information arrives, it updates progress and the incident report so teams spend less time tracking and consolidating work.

Health checks that come back critical now link to an existing incident or open a new one, attaching alerts only when they match the risk. This release also adds model selection and goal conditions to automations, and bulk download on the knowledge page.

This release unifies how tool-call evidence references look and click across chat messages, reports, the right panel, and IM notifications like Lark. It also fixes several Lark document integration permission and configuration issues, and polishes interactions in app management, topology, and document downloads.

This release adds LLM-assisted dashboard query generation, and generated dashboards can now be executed as a whole or by group within a task to cut down repeated tool calls and speed up analysis. It also strengthens JumpServer integration permissions and result retrieval, and fixes issues with Feishu Docs installation, knowledge naming conflicts, and custom MCP calls.

This release adds collaborative incident response, allowing teams to run multiple investigation tasks around the same incident. Incident details bring impact scope, linked alerts, findings, reports, and cleaned activity records into one place so responders can take over quickly, avoid repeated coordination, and track progress throughout the response.

This release focuses on Markdown/HTML PDF export, workspace-level Feature Flags scope, and localization consistency improvements. It also fixes key stability issues, including entity-creation black screens, blank generated HTML pages, and JumpServer permanent-token validation failures.

Castrel can now generate topology knowledge from data sources such as APM and CMDB, showing service and infrastructure relationships, and you can edit it by hand. Users can inspect the topology directly, and Castrel uses it as shared context when handling alerts and investigating incidents.

This release lets you choose an AI model and thinking intensity for each conversation. A WeCom app and Custom Alerts expand collaboration and alert handling, while chat attachments, documents, and knowledge outputs are stored together.

This release introduces the System Knowledge Completion automation task: it scans connected resources for knowledge blind spots, guides teams through structured questionnaires to fill in missing business context, and stores verified knowledge as long-term resource associations. MCP connector support is also expanded across dashboards, CI, and sub-agents, and root cause analysis reports now present multiple candidate root causes.

This release introduces Unified Alert Management, organizing multi-source alerts around incidents so operations teams no longer need to manually piece together clues across platforms. The Prometheus integration adds a one-click toggle to start or stop alert syncing on demand.

This release extends the /SLO command to generate an observability dashboard in one click, orchestrating panels from golden signals and SLO definitions for a unified troubleshooting view. It also optimizes session context building and Skill reading, and fixes several scheduled automation display and health inspection issues.

This release adds a /SLO command that automatically generates SLO documents, brings executable SQL and table inspection to the MySQL and PostgreSQL connectors, and adds a log index query tool with data exploration Skills.

This release unifies loop tasks and automation into a single system, letting tasks run bound to a session or independently. It also enhances search dialog filtering, web citations, and best practices for several connectors, and fixes multiple document and knowledge panel issues.

This release adds an autoload option for knowledge to load full content on demand, streamlines the health inspection command and display structure, and improves best practices and diagnostics for several observability connectors.

This release unifies knowledge base triggering into resource query tools, expanding knowledge automatically by scope. It also adds the DOCP integration with host and CMDB relation queries, and refines the document and knowledge detail page experience.

This release gives the Agent two ways to keep watch on its own. `/loop` turns any command into a recurring scheduled task, and `/goal` sets a completion condition and drives the conversation until it's met — alongside the new `/deploy` and `/health` commands for more ops coverage.

This release introduces per-run Agent permission policies for precise control over what the Agent can do, and adds theme switching with account-level appearance persistence.

This release comprehensively redesigns the Chat interaction experience, introduces interrupt-style AI questioning with conversation steering, and adds application initialization, dashboard generation, and data traceability for more efficient and trustworthy troubleshooting.

This release optimizes the Connector info API with new Skills query support, completes plugin migration for remaining Connectors, and fixes multiple chat session state issues.

This release adds Skills implementations for multiple integration plugins, optimizes the Incident Studio and Automation modules, and introduces an Enterprise tier for SaaS along with batch application management.

This release introduces the brand-new Incident Studio for event-driven investigation workflows, alongside a comprehensive redesign of the Application module with health inspection capabilities. We also unified the AI input experience and completed a large-scale plugin architecture migration.

This release adds Knowledge Canvas detail display and connector data query visualization to Chat sessions, while optimizing Proxy MCP tool execution and introducing Docker-based deployment support.

This release introduces comprehensive task archive and deletion capabilities, with quick archiving from the sidebar and a dedicated archived task management panel. We also added main interaction channel with multi-notification channel support, along with 2 new Connector integrations.

This release focuses on a comprehensive menu redesign .We enhanced Sandbox parsing support for complex file types and optimized the task list display logic and expanded support for core infrastructure components (MySQL, JVM, Redis) and WPS through new Skills and integrations.

This release focuses on two core experience upgrades — Demo Workspace onboarding and enhanced @ mention capabilities. The Demo Workspace provides new users with a seamless trial loop, while the strengthened Connector object referencing further shortens the analysis path from "selecting an object" to "reaching a conclusion."

This update focuses on China Region commercialization, featuring a comprehensive upgrade of the registration experience and the initial launch of the billing system. It also includes optimizations and bug fixes across various scenarios, such as chat document management, Feishu integration, and agent configuration.

This release focuses on a major Sandbox execution upgrade, new integration channels via Custom Sub-Agents and WeChat Clawbot, and stronger platform stability with Castrel Proxy permission controls and automation quota management. Several long-standing UX issues are also resolved.

This release focuses on two core capability areas: automation and Castrel Proxy. Automation now delivers a stronger scheduled-execution and notification loop, while Castrel Proxy adds a new proxy mode for private-network access.

This release focused on a document-first knowledge model refactor and expanded platform capabilities, including broader connector coverage and marketplace-related foundations.

v1.2.0 delivered dark/light theme support, sub-agent improvements in troubleshooting flows, and stronger connector/tooling compatibility.

v1.1.0 focused on @knowledge in chat, proxy execution-mode upgrades, and connector/channel reliability improvements.

v1.0.0 established the first stable SaaS baseline for Castrel, including core chat flow, connector management, and initial Castrel Proxy capabilities.