“What breaks if I change this component?” sounds like a code question. The answer may also depend on who owns the service, which pipeline ships it and where the latest change actually ran. An agent that sees only files starts with part of the story.

GitLab announced Orbit's general availability on October 29. The product organizes software lifecycle relationships into a graph that people and agents can query. Availability begins on GitLab.com for Premium and Ultimate customers, with other deployment options following over the coming weeks.

The useful context lies between records

Orbit's documentation describes queries across code and work data, filtered by the user's permissions. Agents can connect through the CLI or MCP, a protocol for integrating tools with AI systems.

Consider a shared-library migration. Repository references help locate callers, but do not by themselves establish which projects still use an old version, who can approve changes or whether a fix reached production. Organized context can reduce the separate investigations needed to assemble that answer.

This is a practical interpretation of the proposal, not a MnzAI Labs product test. It illustrates why adding documents to a prompt does not always solve the problem. Teams need current, verifiable relationships between records, rather than simply more text in a context window.

A better-contextualized answer still needs evidence. A useful result should let the reviewer trace the relationship queried and check whether it reflects the current state of the work.

Fewer tokens do not automatically mean a lower bill

GitLab publishes results from its own and partners' testing to support efficiency claims. Those are vendor-attributed findings, not a savings guarantee for every team.

Reducing tokens on a task is different from lowering total operating cost. Graph-query charges, answer quality and review time all matter. If an inexpensive query produces an incomplete conclusion, subsequent work can consume the apparent saving.

A local evaluation can use questions the team already knows how to answer and compare the effort required to reach a verifiable result. Holding tasks, permissions and acceptance criteria consistent helps establish what the integration adds. Measuring token consumption alone would miss an important part of the outcome.

Permissions are part of the context

A shared graph also has to respect who may retrieve each record. The documentation makes permission filtering part of the product's contract. That deserves attention when agents access information across projects.

A proof of concept can ask the same question using accounts with different access and check whether the answers preserve that distinction. This is a suggested validation, not an identified Orbit vulnerability.

General availability makes it possible to test the proposal against real work. Adoption becomes easier to judge when the question goes beyond whether the agent answered faster: did it show correct relationships, within the allowed access, at a cost that justified the result?

Sources: GitLab's announcement and Orbit documentation.