Changelog

Coding agents can now build Tracecat automations. The Tracecat automation skill ships open source in the main repo: load it into Claude Code or any skills-capable agent connected over Tracecat MCP, and it authors workflow YAML, designs tables, validates definitions, and debugs executions the way our own team does.

The skill covers the full authoring loop, from discovering the workspace and reading the DSL reference to running workflows and inspecting failures. A companion skill handles Slack-facing automations.

Agents can now load skills: reusable instructions that capture how your team enriches alerts, escalates incidents, or works with specific tools. An agent loads a skill on demand when the task calls for it, so the same preset agent follows your team's procedures.

Build and manage skills in the new skills studio.

Dropdowns, custom fields, and metrics now render directly in case rows, so you can set severity or priority from the list without opening the case. Closure requirements let you mark fields that must be filled in before a case can close.

Custom fields gain long-text and URL types, and bulk actions now cover commenting, appending descriptions, and setting dropdowns across many cases at once.

The chat panel from cases now opens on workflows, tables, and agents. Each conversation is a full agent session with real tools and your workspace in context, replacing the scripted copilot that previously backed these surfaces.

One agent runtime now sits behind chat everywhere in the product.

MCP servers can now run on Tracecat's infrastructure. Connect agents to third-party tools through a hosted server instead of deploying and operating the connector yourself.

Jira is the first hosted integration, secured with OAuth 2.1. More servers are following.

Tracecat now runs as an MCP server. Connect Claude, Codex, or any MCP client to your workspace and work with workflows, cases, and tables from the assistant you already use.

The server exposes the same operations as the UI, so anything you can do in the product, a connected assistant can do too.

Tracecat Cloud is live. The platform is now a multi-tenant service with organizations, workspaces, and entitlements built in, running on a rebuilt execution engine with pure Temporal orchestration and large payload support.

Major surfaces were rebuilt along the way. Tables gain direct cell editing and Markdown text cells, case management tables are revamped, and every tool secret and OAuth connection now lives in a single integrations view.

Agents can now pause for human sign-off before running sensitive actions. The new approvals agent stops at designated steps and waits until someone approves or denies the action before continuing.

A centralized approvals dashboard lists every pending request, so reviewers can approve or deny across all runs from one place.

Workflows can now be backed by a git repository. Git sync, available behind a feature flag, stores workflow definitions in a repo you choose, so edits land as commits and the builder stays in sync with the branch you point it at.

Changes can be reviewed as pull requests, and the commit history records who changed what.

Every case now has an AI chat panel. Ask questions about the incident, run enrichment, and work the case in conversation with an agent that has the case in context.

Case tags ship in the same release, alongside a streamlined case page.

Pull actions from your own remote repositories and write new template actions in plain YAML. Your team's integrations install alongside the built-ins, with no fork of Tracecat required.

Underneath is the Tracecat Registry, an open actions registry behind the whole platform. Browse and manage every action and repository from the UI, and attach secret keys to actions that need credentials.