GitLab expands AWS integration for agentic AI workflows

GitLab has expanded its integration with Amazon Web Services, allowing joint customers to route GitLab Duo Agent Platform inference through Amazon Bedrock.

The move targets companies already using AWS for generative AI and software delivery. Customers can use Amazon Bedrock models, AWS Identity and Access Management policies, and existing AWS spending commitments without setting up a separate AI stack.

The integration extends GitLab's Duo Agent Platform, which supports AI-driven DevSecOps tasks such as code generation, merge request handling, and pipeline support. Under the new setup, organisations can direct inference through Amazon Bedrock accounts they already manage instead of creating new model endpoints or separate billing relationships.

For companies that buy GitLab through AWS Marketplace, GitLab Credits used by the Duo Agent Platform will count toward existing AWS spending commitments. This allows customers to expand their use of agentic AI within current commercial arrangements rather than negotiate new supplier contracts.

Governance layer

The integration adds workflow governance on top of Amazon Bedrock's model controls. Because the platform already records merge requests, pipelines, and security findings, administrators can apply policies to govern which models AI agents may use and capture audit logs for agent actions alongside the code produced.

Teams can also establish approved configurations so agents operate consistently across projects and regions. That is likely to appeal to larger enterprises trying to bring AI into software development while maintaining oversight and internal controls.

Another part of the announcement builds on GitLab's Bring Your Own Model option for self-managed customers. This allows teams to connect a self-hosted AI Gateway directly to Amazon Bedrock within their own AWS environments.

As a result, source code and inference traffic can remain within a customer's network boundary. Customers with fine-tuned models on Amazon Bedrock can use those directly, while others can choose GitLab-managed models, including Anthropic Claude models available through Amazon Bedrock.

Enterprise focus

The announcement reflects a broader push by software vendors to tie AI services more closely to cloud environments enterprise customers already use. Rather than asking customers to move data into separate systems, vendors are increasingly emphasising deployment models that fit existing cloud controls, procurement routes, and governance structures.

That matters as businesses look to deploy AI agents more widely across software development. Greater automation in coding and testing can increase the volume of changes, reviews, and pipeline activity, which raises the need for audit trails and policy enforcement.

Rahul Pathak, Vice President, Data & AI GTM, AWS, said the integration is designed to reduce that friction for AWS customers. "GitLab Duo Agent Platform on Amazon Bedrock means customers can deploy agentic AI for software development without standing up new infrastructure, negotiating new contracts, or rethinking their security posture," he said. "It runs inside their existing AWS environment - same IAM policies, same compliance controls, same spending commitments. That's how AI adoption should work."

GitLab also presented the move as a way to help companies avoid duplicating technical and commercial arrangements when adopting AI tools. Many customers want to use AI within software delivery without adding a second layer of infrastructure and governance, it said.

"Most enterprise leaders I talk to want to adopt agentic AI without building a second stack next to the cloud environment they already use," said Manav Khurana, Chief Product and Marketing Officer, GitLab. "For AWS customers, this integration makes that possible. GitLab Duo Agent Platform runs through Amazon Bedrock accounts they already manage, governed by policies they already enforce, and funded by commitments they've already made. AI adoption scales when it fits into decisions already made, not when it asks teams to make new ones."

One customer cited the ability to keep traffic and code within AWS as a key factor. "As an organisation standardising on Amazon Bedrock for generative AI, GitLab Duo Agent Platform with Bring Your Own Model (BYOM) will let us apply AI across our software delivery lifecycle while keeping inference traffic and code within our AWS environment," said Yoshiki Matsuda, Chief Operating Officer, Fixstars Corporation. "At Fixstars, we view AI as central to modern software development, and our own Fixstars AIBooster has already driven significant improvements in both the quality and speed of our application development. What drew us to the Bring Your Own Model approach is GitLab's ability to orchestrate, secure, and govern workflows on top of the Bedrock investments and controls we already trust, without requiring a separate AI stack."