GitLab Orbit reaches general availability, promising fewer agent retries and tokens

GitLab Orbit, a live context graph covering code, merge requests, pipelines, vulnerabilities and ownership, is now generally available. According to GitLab, agents connected to Orbit needed up to 45x fewer retries and used 4.5x fewer tokens on the same tasks with the same model.
- GitLab Orbit, a context graph for code, merge requests, pipelines, vulnerabilities and ownership, is now generally available.
- GitLab says agents using Orbit needed up to 45x fewer retries and 4.5x fewer tokens on the same tasks with the same model.
- Compare the Market's Orbit-grounded reviewer matched about 70% of expected inline comments across 79 merge requests, versus about 58% for RAG.
- Each query pack costs $30 in GitLab Credits per user per month and adds 1,000 queries to a shared pool.
- Available to Premium and Ultimate customers, starting with GitLab.com, with Self-Managed and Dedicated to follow in coming weeks.
A graph of the whole software lifecycle
GitLab argues that the main limit for coding agents is not the model but context. In its announcement, the company says agents guess because code is all they can see. They cannot see the merge request that last changed it, the pipeline that ran it, the vulnerability it carries, or the people who wrote and reviewed it. In large monorepos with tens of thousands of files, GitLab says, agents keep pinging the backend for information, and every round trip costs time and tokens until the agent eventually gives up.
Orbit is GitLab's answer. It is described as a live, queryable view of code, merge requests, work items, pipelines, vulnerabilities and ownership, with relationships updated continuously as teams ship. An agent such as Claude Code, connected through MCP, queries the graph instead of crawling files. According to GitLab, the first response returns real callers, recent changes and owners, so follow-up prompts refine the answer rather than correct it.
Test results and customer feedback
The headline figures come from GitLab's own testing: up to 45x fewer retries and 4.5x fewer tokens on identical tasks with the same model. GitLab says the cost of an undisciplined agent session is dominated by retransmitting context that did not change between turns, a finding it attributes to a recent analysis by a leading research and advisory firm. Because the graph is shared, GitLab adds, the cost does not repeat per agent.
Compare the Market, a UK price comparison platform, tested four context approaches for an internal AI code reviewer across 79 real merge requests. The Orbit-grounded reviewer matched about 70% of expected inline comments, while RAG managed about 58%, which GitLab notes was worse than no added context at all. Ryan Harvey, Head of AI Engineering at Compare the Market, said: "Better comment placement, better summaries of what actually changed."
Since Orbit entered beta in June, design partners have used it on real codebases and incidents. GitLab cites several examples:
- One partner answered a security-coverage question across 34 million entities in seven minutes. Through the API, GitLab says, the same question would have taken half a day.
- A financial services company traced a build-system outage to a CI job failing 50% of the time, a pattern nothing else had correlated.
- Another partner finished a multi-repository upgrade by asking which projects were still on the old version, surfacing the last stragglers in one query.
How it connects and what it requires
GitLab says Orbit draws on data already in GitLab, so there is no connector to configure, no ETL pipeline to maintain and no second system of record. It mirrors nested groups, roles and security-domain access, so permissions stay intact. The company also says Orbit is built to extend to non-GitLab data sources.
Agents on the GitLab Duo Agent Platform query Orbit natively. Claude Code, Codex, Cursor, OpenCode, Gemini CLI, Antigravity and other MCP-compatible agents connect through MCP or the GitLab CLI, using the commands glab auth login and glab orbit setup. Custom agents connect through REST, and engineers can query the same graph through the Data Explorer.
Orbit indexes every push, merge and deploy as it happens. At GitLab's own scale, the company says, that means more than 1.5 million projects and 8TB of graph data in under 2 hours. Indexing runs as a separate service beside GitLab, so heavy agent use does not slow GitLab itself, according to the company.
Pricing and availability
Orbit is sold as query packs that form one shared pool for an organization, drawn on by humans and agents alike, with no seats to assign. Each pack costs $30 in GitLab Credits per user per month and adds 1,000 queries to the pool. Usage beyond the pool is billed at 33 queries per GitLab Credit, about $0.03 per query. GitLab says promotional pricing is available this quarter for early customers.
Orbit is generally available for Premium and Ultimate customers, starting with GitLab.com today and reaching Self-Managed and GitLab Dedicated over the coming weeks. Customers can try it free for 30 days through their GitLab account team.
Use in security work
In a companion post, GitLab positions Orbit for finding the blast radius of a vulnerability. Using the recently disclosed PHP2Shell flaws as an example, it argues that scanners show different pieces of the picture, while teams still need to know which services are affected, who owns them and which pipelines deploy them. GitLab says Orbit replaces that manual work with a single query that returns an answer in minutes.


