The Atlassian estate counts ten members and 27 APIs, and leaves out Trello, Loom and 183 APIs on its own parent profile

The Atlassian estate counts ten members and 27 APIs, and leaves out Trello, Loom and 183 APIs on its own parent profile

The Atlassian estate says Atlassian’s API surface spans 10 provider profiles and 27 APIs. The estate rating is 30.8, thin, and the rated members run from 76.2 down to 0.0. Both numbers understate how much Atlassian publishes and overstate how coherent it is.

What the estate counts

Member Relationship Kin Score Agent Readiness
Jira product 76.2 exemplar 65.6 agent-ready
Atlassian Compass product 72.6 exemplar 48.0 agent-ready
Confluence product 70.1 exemplar 54.4 agent-ready
Bitbucket acquisition 70.0 exemplar 34.2 agent-ready
OpsGenie acquisition 48.1 developing 28.3 agent-aware
Bitbucket Pipelines product 45.7 developing 31.2 agent-ready
Statuspage product 33.6 thin 18.3 agent-aware
Optic acquisition 21.0 emerging 2.5 human-only
Wikidocs acquisition 3.4 minimal 0.0 human-only
HipChat product 0.0 minimal 0.0 human-only

The composite average of 36.3 includes HipChat, which is marked defunct and scored zero by rule, and Wikidocs, an acquisition with no API. By my arithmetic, the eight live members average 54.7. The estate as published describes Atlassian’s history more than the surface a developer can use today.

What it leaves out

The 27 APIs come from each member’s declared api_count, and the parent is not a member row. The Atlassian provider profile carries 183 API pages of its own at 67.5, exemplar. They include 22 pages named for Jira (10 of them Jira Service Management), 13 for Bitbucket, and one each for Trello, Compass and Statuspage.

Three companies are missing from the membership altogether. Trello describes itself as “from Atlassian,” carries the Atlassian tag, and scores 75.7, exemplar, which would make it the second-best member. Loom, “now part of Atlassian,” scores 21.2. Halp, acquired in 2020 and folded into Jira Service Management, scores 3.4. Jira’s record declares x-parent: atlassian. None of these three declares a parent, so the estate cannot see them.

What fragmentation costs an integrator

Say you are building on Bitbucket. You will find it in three places: 13 pages on the parent profile, 23 on the Bitbucket profile, and 23 more on Bitbucket Pipelines, which is its own subfamily. Jira is split the same way, across the parent and a Jira profile with 132 pages. That Jira profile is also where the Atlassian Remote MCP Server is listed. The parent’s own MCP entry has no description.

Two surfaces that were once duplicates, atlassian-jira and atlassian-confluence, have already been merged into Jira and Confluence and now redirect. That was the right fix. The parent profile is the next place to apply it, because an agent resolving “Atlassian” today reaches a 46.6 agent-ready parent, while Jira, a single click away, is 65.6.

What would move it

On our side, we should declare the parent on Trello, Loom and Halp, and reconcile the parent’s Jira and Bitbucket pages with the product profiles they duplicate. On Atlassian’s side, publishing one catalog across products, such as an APIs.json or a well-known API catalog that names every product surface, would let any crawler rebuild this estate without guessing.

See the estate at apis.io/estates/atlassian/.

← a16z's twelve agent-native companies mostly move money, and OpenAI is not one of them
AWS ships 38 healthcare agent skills with a benchmark, and Bedrock's own record lights none →