GitLab Orbit comes to Google Antigravity, giving AI agents context on the entire project

GitLab has integrated Orbit, its lifecycle context graph, into Google's Antigravity platform. Developers can now install the integration directly from the MCP Store and give agents structured access to data from their GitLab instance.

What is GitLab Orbit

Orbit indexes an entire GitLab instance and builds a knowledge graph of the relationships between groups, projects, users, issues, merge requests, pipelines, vulnerabilities, and source code. The graph is exposed through two MCP tools: `query_graph` runs structured queries, and `get_graph_schema` returns the available node types, properties, and relationships.

Before Orbit, Antigravity agents could see only open files and the terminal. They didn't know which services depended on the code being changed, couldn't find similar vulnerabilities in other projects, and had no idea who had historically worked on those files. Development lifecycle context stayed locked inside the DevSecOps platform, reachable only through custom scripts or by copying and pasting between tabs.

What the integration delivers

With Orbit connected, an agent can answer questions that used to require multi-step manual searching:

• Which projects depend on this module, and will the change break them?
• Are there unresolved vulnerabilities in the project?
• Who should review this merge request, based on review history and file ownership?
• Which projects cause pipeline failures most often in the group?

The agent writes a query in Orbit's JSON DSL and gets back a typed response, with no switching between browser tabs. In early internal tests, agents connected to Orbit answered 11 times faster, used 4.5 times fewer tokens, and hallucinated 45 times less often.

Three key scenarios

Blast radius analysis. Before refactoring a shared authentication library, an engineer asks the agent: what depends on this module, which open merge requests touch these files, and who owns them? The agent sends a single query to the graph and returns all three answers at once. The engineer sees which issues the refactor will run into before making the first edit.

Onboarding to a new service. A developer returning to unfamiliar code asks the agent to show dependencies, entry points, and this week's merge requests. The agent builds a "Walkthrough Artifact" — a map of the service's current state, not the outdated wiki that usually greets newcomers. Orbit reindexes within minutes of any change, so the map is always fresh.

Dependency map. A tech lead requests the dependency structure for a group of services and asks the agent to render it as an architecture diagram. The graph is live, so the diagram reflects the current state rather than a schema nobody has maintained by hand in a long time. For a narrower slice, such as only services with open vulnerabilities, a follow-up query and a regenerated diagram are all it takes.

Installation and availability

Orbit installs in a few clicks from the MCP Store in Antigravity settings: open the MCP section, click "Add MCP", select GitLab Orbit, and authenticate with GitLab. No config files or terminal required. Once installed, Orbit's tools are automatically available to every agent in the workspace.

Orbit is the same engine that supplies context to the GitLab Duo Agent Platform. Platform engineering teams get a single knowledge graph for agents both inside GitLab and in Antigravity, without a separate context pipeline. Supported languages are Ruby, Java, Kotlin, Python, TypeScript, JavaScript, Rust, and C#. The integration is available on GitLab Premium and Ultimate tiers on GitLab.com.

What it means

GitLab Orbit's arrival in Google Antigravity moves AI agents closer to working with real repository context instead of hallucinating because they lack it. Instead of custom scripts and copy-paste, developers get a structured knowledge graph right in their working environment, and agents start giving answers that account for the full complexity of a live project.