Jira and Confluence

One Atlassian Cloud site that answers Jira, Jira Software, Jira Service Management and Confluence REST, plus Atlassian's MCP tools.

Surfaces

Kind	Path	Notes
REST	`/rest/api/3`	Jira platform REST, also `/rest/api/2` and `/rest/api/latest`
REST	`/rest/agile/1.0`	Jira Software boards, sprints and backlogs
REST	`/rest/servicedeskapi`	Jira Service Management
REST	`/wiki/rest/api`	Confluence v1, and Confluence v2 under `/wiki/api/v2`
MCP	`/mcp`	Atlassian MCP tools for Jira and Confluence

API versions: Jira REST v3, v2, latest, Jira Software agile 1.0, Confluence REST v1 and v2

Each product's pagination style, including nextPageToken on /search/jql and cursor-only Confluence v2
Each product's error envelope, with 400 for malformed JQL/CQL, 404 for unknown ids and 409 for a stale Confluence page version
A JQL and CQL subset, including the history predicates CHANGED and WAS
The 19 built-in custom-field types the administration API accepts
OAuth scopes, product access, permission schemes, issue security levels and Confluence page restrictions decide whether a call is allowed
Blog posts, page and blueprint templates, and content properties

JQL sub-queries (answered 400 with the vendor envelope)
Sending notifications; notification schemes are served but nothing is sent
Cascading selects and app-provided custom-field types
Confluence custom content beyond an empty collection, and space permissions
Rovo search (searchAtlassian/fetchAtlassian), Teamwork Graph, Compass, Bitbucket and Jira Service Management MCP tools

Scenario	Difference from Jira and Confluence	In your tests
Text search (`~`, CQL `text`)	Whole-word matching with simple plural and `-ing` folding, phrases and wildcards; no relevance ranking, synonyms or language analysis. `/search/approximate-count` returns the exact count.	Assert that an issue or page is found, not on its order or on fuzzy matches.
Authentication	Open mode accepts any credential; strict mode only checks the token came from the bundled IdP. Authorization is still evaluated.	Test what a credential may do, not whether a bad credential is rejected by Atlassian identity.
User identifiers	Users carry the Cloud `accountId` and also the Server/Data Center `name` and `key`, accepted wherever a user is named.	Do not assert that `name` or `key` are absent; use `accountId` for Cloud behaviour.
Dates	Generated activity is shifted forward so relative-date queries match.	Assert with relative JQL (`created >= -7d`), not on absolute dates.
Rate limits	Nothing throttles until a fault is injected.	Inject `throttle` or `fail_next` to test backoff.
Page size	`maxResults` clamps to 100 on Jira and `limit` to 200 on Confluence v1, as Atlassian does, rather than answering 400.	Follow pagination; do not assume a large page size is honoured.

Key	What the client sees
`throttle`	429 with `Retry-After`, `X-RateLimit-*` headers and `RateLimit-Reason` `jira-burst-based`, in the Jira or Confluence envelope
`fail_next`	The next N calls fail as `throttle` does, then clear
`quota_limit`	429 with `RateLimit-Reason` `jira-quota-global-based`
`error_rate`	500 in the product's error envelope for the given fraction of calls
`latency_ms`	Fixed added latency on every data-plane request

The hosted data plane is read-only: a write is refused with 403 writes_disabled. A SOQL, GraphQL or search POST is a read and is answered.
Faults are injected through POST /_admin/faults on a simulator you run yourself; POST /_admin/faults/reset clears them.
Rate limits do not happen on their own unless a page says so. Use fault injection to exercise a client's backoff.
Distributions come from aggregated metadata sketches of data Eon backs up; no customer records; all Era data is simulated.