GitLab adds goal-driven flows and Slack agent to cut waits between dev stages

The bottleneck has moved downstream

According to GitLab, AI coding tools have changed how fast software gets written, but not how fast it ships. As code generation speeds up, the company says, the slowdown shifts into verify, review and ship, stages that still run on human time. Agents can write fixes in minutes, yet delivery is held back by wait times and manual handoffs.

GitLab lists reviewer signoffs, pipeline alerts, scan triage and step-by-step prompting as typical obstacles for DevOps teams trying to deliver on time. The new capabilities are meant to address that delay. All three run under the user's identity, policy and evidence chain, so governance stays with the work, the company says.

Goal-driven flows, Slack and Custom Flows

Goal-driven flows give agents the finish line instead of the next step. A user describes the objective and what done looks like, and the flow keeps working until it gets there. Review, tests, security scans and approvals run as one connected sequence. A flow is started with the /goal command, for example: /goal Fix the checkout timeout. Continue until tests pass and CI is green. According to GitLab, a separate model checks the work against the goal at each step, and users can stop or redirect a run at any point. Flows can be started in Duo CLI, in headless mode with duo run --goal, and in Agentic Chat.

GitLab for Slack lets teams start work without leaving Slack. Mentioning @GitLab in a channel or thread and describing a need, such as "fix the failing pipeline," starts a flow. Duo streams progress back into the thread. The @GitLab mention can also be used to ask questions about a codebase or to create an issue from the conversation.

Custom Flows chain agents together to match a team's existing process. A flow is defined once, started from GitLab events such as a mention, an assignment or a pipeline, and run in CI/CD pipelines. GitLab gives the example of a self-healing pipeline that analyzes a failed test, generates a fix, commits it and notifies the team, with human-in-the-loop checkpoints at sensitive steps.

Availability and requirements

Goal-driven flows are available on GitLab.com, GitLab Self-Managed and GitLab Dedicated, on the Premium and Ultimate tiers. They require GitLab 19.3 or later and Duo CLI 9.17.0 or later. GitLab recommends starting with goal-driven flows and points to the Duo CLI slash commands documentation for the /goal command.

Custom Flows are generally available. The MCP server and GitLab for Slack will be generally available later this month, according to the post. Foundational agents and flows already cover plan through analyze stages in the GitLab UI, the IDE and Duo CLI. Teams can enable more from the AI Catalog, bring in external agents, and connect tools through the MCP client and the MCP server.

What it means for teams

For teams on Premium or Ultimate, the changes shift agent use from single prompts toward outcome-based runs that span several stages. GitLab says users keep control through the ability to stop or redirect a flow and through checkpoints in Custom Flows. New users can start a free trial of GitLab Duo Agent Platform, while existing Premium and Ultimate customers can turn it on and use the GitLab Credits included with their subscription.