Atlassian

Adds a relationship between two entities in the Teamwork Graph (e.g. linking two Jira work items, marking one as blocking another, attaching a remote link, or connecting a Jira work item to an Atlas project or goal).

addGraphContext

Add a comment to a Jira issue, or edit an existing one: omit commentId to add, or pass one from listJiraIssueComments to edit. Returns the commentId, needed to edit or delete it later. When no file was appended through inlineFileId, it also returns the echoed body and its format (appliedContentFormat).

addOrEditJiraIssueComment

Open Atlassian Home to see pull requests and recent work.

atlassianHome

Create a new Confluence page, blog post, live doc, whiteboard, database, embed, smart link, or folder. Before choosing body.format or authoring a body, call getContentFormatGuide through execute with name=getContentFormatGuide and inputs={toolName}. Use createConfluencePage for pages, blog posts, and live docs; createConfluenceWhiteboard for whiteboards; or createConfluenceDatabase for databases. Space instructions describe how to author content in a Confluence space. Apply relevant guidance to the content you create or update. Follow this decision rule: 1. Before authoring, call getConfluenceSpace for the target space unless getConfluenceContent returned metadata.hasSpaceInstructions=false for content in that space. If getConfluenceSpace already returned for the same cloudId and space during the current task, reuse its spaceInstructions and do not call it again. 2. When the value is true or unknown and instructions have not been loaded, you must call getConfluenceSpace once before authoring through execute with name=getConfluenceSpace, top-level cloudId, and inputs={spaceIdOrKey}. 3. Apply any returned spaceInstructions. If a successful response omits them, no instructions are configured; author with defaults and do not block. After a successful getConfluenceSpace call, do not make a separate getConfluenceSpaceInstructions call.

createConfluenceContent

Create a new Jira issue. Custom fields go in additional_fields: keyed by field name, values are resolved automatically, so no field lookup is needed first (user fields take accountIds); keyed by customfield_* ID, pass the value in Jira's format. A rejected create returns a repairHint and names the field at fault where there is one, with its allowedValues when known, so retry from that. If sprint assignment fails, the response still has the new issue's key and sprintAssignment.assigned: false — do not create it again.

createJiraIssue

Find Atlassian operations that are not already in your tool list. Searches ~310 operations across Jira, Confluence, Compass, Bitbucket, Focus, Goals, Projects, Loom, Talent, Teams, Capacity-planning, Jira Align, Artifacts. Returns up to 5 matches with the exact name and inputs needed to call the matching execute-family tool (executeRead, executeWrite, or executeDestructive), then a briefer name-and-summary list of further matches — re-query with narrower terms to get one of those in full. Read all 5 before choosing. Only use an operation name from discover results or one already in your tool list; never guess, assume, or invent a name — when unsure, call discover first.

discover

Edit an existing Jira issue. Custom fields go in additional_fields: keyed by field name, values are resolved automatically (user fields take accountIds); keyed by customfield_* ID, pass the value in Jira's format. Every field is set, not appended: send the full list for labels, components and other multi-value fields. Change status or move to a sprint with transitionJiraIssue, and convert Task↔Sub-task with convertJiraIssueHierarchy. Empty description or environment with "". When editing a body that holds inline media, use contentFormat: html, or the media is dropped. Reorder on a board with rankBeforeIssue or rankAfterIssue; fields are saved first, so a failed reorder leaves the edit applied and returns ranking.ranked: false. The response shows only getJiraIssue's default fields, so other fields you set are not echoed back.

editJiraIssue

Run a destructive Atlassian operation by name (deletes or irreversibly changes data) across all Atlassian products. Only use an operation name from discover results or one already in your tool list; never guess, assume, or invent a name — when unsure, call discover first. If the user provides a Jira/Confluence URL, pass it directly as cloudId — no need to call getAccessibleAtlassianResources. Pass all inputs as a flat object — the server routes each to the right location (path, query, or body) automatically. Use responseFields for field filtering, e.g. ["key","fields.summary","fields.status.name"]. Use view to select a named output preset (e.g. "compact", "evidence", "full") when the operation supports views.

executeDestructive

Run a read-only Atlassian operation by name across all Atlassian products. Only use an operation name from discover results or one already in your tool list; never guess, assume, or invent a name — when unsure, call discover first. If the user provides a Jira/Confluence URL, pass it directly as cloudId — no need to call getAccessibleAtlassianResources. Pass all inputs as a flat object — the server routes each to the right location (path, query, or body) automatically. Use responseFields for field filtering, e.g. ["key","fields.summary","fields.status.name"]. Use view to select a named output preset (e.g. "compact", "evidence", "full") when the operation supports views.

executeRead

Run a write Atlassian operation by name (creates or updates data, non-destructive) across all Atlassian products. Only use an operation name from discover results or one already in your tool list; never guess, assume, or invent a name — when unsure, call discover first. If the user provides a Jira/Confluence URL, pass it directly as cloudId — no need to call getAccessibleAtlassianResources. Pass all inputs as a flat object — the server routes each to the right location (path, query, or body) automatically. Use responseFields for field filtering, e.g. ["key","fields.summary","fields.status.name"]. Use view to select a named output preset (e.g. "compact", "evidence", "full") when the operation supports views.

executeWrite