LoginSign Up
← Back to all sparks
GitLab logo

GitLab

DEVOPSCOLLABINFRA · APIS
Velocity6.3

DevOps platform

gitlab.com ↗

GitLab's week is guides: Jira tickets to merge requests, and cleaning up leaked agent secrets.

ai agentsduo agent platformsecretsintegrationsknowledge graph
◆Current state
Orbit, GitLab's queryable graph of code, merge requests, pipelines and owners, went GA on October 29 and is still the last product change. The two newest posts are guides built on existing features. One wires a custom Jira agent into Duo Agent Platform so a ticket goes to Developer Flow and comes back as a merge request. The other shows how to block, find and remove the Claude Code settings file that leaked tokens into public repositories.
◆Where it's heading
GitLab is positioning itself as the place where coding agents are connected and supervised: Orbit gives agents context, Duo Agent Platform connects them to outside systems through MCP, and push rules and secret detection catch what they leave behind. The posts since GA explain that stack to buyers rather than extend it.
◆Prediction
Expect more Duo Agent Platform integrations with outside tools, likely moving the Atlassian MCP server in the AI Catalog past experiment status; the entries give no date.

◆Recent moves

  1. 2d ago

    Turn Jira tickets into GitLab merge requests with Duo Agent Platform

    A tutorial that combines the existing Jira integration, a custom agent and the experimental Atlassian MCP server so a Jira ticket becomes a merge request through Developer Flow. It shows the agent platform reaching into outside tools, but nothing new ships.

    View source ↗
  2. 3d ago

    How to prevent leaks from Claude Code's 'Don't ask again' file

    A response guide to Claude Code's local settings file leaking tokens into public repositories, using .gitignore templates, push rules and secret detection. Like the earlier CVE guides, it applies existing controls to an outside incident.

    View source ↗
  3. 5d ago

    How to design a decision-grade pilot for agentic AI

    A framework for agentic AI pilots that asks what evidence would authorize an agent for a specific workflow. It is buyer guidance that supports GitLab's governance pitch, with no product change.

    View source ↗
  4. 8d ago

    The Pentagon wants to know who checks code when AI writes it

    Maps the Pentagon's new rule, which treats AI-written code as unverified until reviewed and traced, to GitLab's existing review, scanning and audit features. It shows GitLab aiming its AI governance story at defense buyers, but nothing ships.

    View source ↗
  5. 12d ago

    10 questions to ask any software vendor about sovereignty

    A buyer's checklist on software sovereignty, tied to GitLab's comment on the EU's proposed Cloud and AI Development Act. It is positioning, not a product change, but it signals GitLab expects data residency and control to become procurement criteria in Europe.

    View source ↗
  6. 16d ago

    GitLab Orbit is GA: Higher agent accuracy, lower token cost

    ⚡ SPARK

    Orbit's general availability is the anchor of GitLab's current arc: the lifecycle data in the platform becomes structured context agents can query. Everything else in this stretch, from security playbooks to governance, builds on it.

    View source ↗