Atlassian Makes Its Forge LLMs API Generally Available
Forge developers can now call hosted Claude models through one SDK while keeping application data inside Atlassian's cloud boundary.
Atlassian made its Forge LLMs API generally available on July 29, moving the hosted model interface beyond its June preview. The release gives developers one API for adding language-model features to Forge applications while Atlassian manages the underlying model infrastructure and keeps application data within its cloud boundary, according to the company's official announcement.
What changed
The API runs on Amazon Bedrock and currently exposes Claude model tiers covering Sonnet, Opus and Haiku. Atlassian says the general-availability release upgrades Sonnet support to Sonnet 5 and makes Opus 4.7 and 4.8 available. Developers call the models through the Forge LLMs SDK rather than managing separate model API keys. Existing Forge developers do not need a separate AI account or a change to their app billing setup.
The data path is a central part of the offer. Atlassian says model calls stay inside the Atlassian platform through Bedrock, so qualifying applications can retain the Runs on Atlassian designation. That addresses a common enterprise review question: whether an app sends customer data to a separately operated AI service.
Adoption and limits
Atlassian reports that more than 100 unique apps using Forge LLMs reached production during the preview period. Twenty were public Marketplace apps serving customers, while more than 350 additional apps were in development. Those are vendor-reported adoption figures, not independently audited usage measurements.
The current model catalogue is also narrower than Bedrock's full range. Atlassian says it plans to expand selection in the future, but the announcement does not set dates or name additional models. General availability therefore confirms support and platform status; it does not guarantee every Bedrock model will become available through Forge.
Why it matters
For Atlassian developers, the release reduces two pieces of integration work: operating a separate model connection and documenting data egress to another AI provider. It also gives Marketplace builders and internal development teams the same supported route for adding summarisation, structured-content generation and workflow assistance inside Jira and Confluence extensions.
The larger platform signal is that Atlassian is treating hosted model access as a standard Forge building block alongside custom Rovo agents and skills. The next test is whether the company broadens model choice while maintaining the same cloud-boundary and governance properties.
Status
Confirmed. Atlassian's official developer announcement supports the release details. Internal confidence is medium because the availability, architecture and adoption figures come from the vendor without independent corroboration.
Sources
Update note: Last reviewed 2026-08-02. We will revise this post if Atlassian changes model availability or the stated data-handling boundary.
Sources
Drafted with AI assistance from source briefs; reviewed for citation completeness and label accuracy.