KickSide Atlassian

Jira and Confluence provider for KickSide.

Auth

Two connection modes resolve to a single authenticated transport, so the service and tool layers never branch on auth:

• OAuth (3LO) — self-declared oauth_provider: atlassian, discovered by kickside/oauth and run through kickside.oauth:base_connector. Tokens are stored and refreshed by the oauth subsystem, keyed by the connection component_id. transport.connect resolves the Atlassian cloudId from /oauth/token/accessible-resources to build the api.atlassian.com/ex base.
• Stored credentials — one kickside.connection:connection binding (provider: atlassian) covers both Jira and Confluence. Cloud uses Basic email:api_token; Server/Data Center uses Bearer PAT. Secrets live in the connection component's private_context.

Layout

• core/ — shared transport (auth + HTTP), site/base-url resolution, types.
• connection/ — shared Atlassian connection binding and live Jira/Confluence resource discovery.
• jira/ — Jira REST client and agent traits/tools.
• confluence/ — Confluence REST client and agent traits/tools.

Planes

• A — agent tools: JiraRead/JiraWrite, ConfluenceRead/ConfluenceWrite over OAuth or credential connections.
• B — source flows: pullable source ports expose Jira issues and Confluence pages. Data Sync selects the connection and source configuration; the shared Automation engine owns scheduling, cursors, rounds, and delivery.

Jira source payload and cursor

The Jira Issues port declares the fields in item.payload as its output schema. Pull envelopes carry item_key, dedup_key, op, and source_version separately. issue_priority contains the Jira priority name; properties.priority retains its original object. Content destinations can use title, content, and content_type directly with a pass-through transform.

Cloud search uses nextPageToken and isLast. Each nonfinal cursor holds the continuation token, the fixed JQL query, and the greatest observed update time. A resumed or replayed page uses that same query. The final cursor clears the page position and starts live polling after the committed update watermark. Server/Data Center search retains startAt pagination with the same fixed-query behavior. JiraRead and the Jira source render ADF descriptions as readable text, preserving inline boundaries and line breaks. JiraRead also renders ADF comment bodies.