Confluence

Read and write Confluence pages, search content, manage labels and comments. Use when user mentions Confluence, wiki, documentation, pages, or knowledge base.

Confluence Cloud

Read and write Confluence Cloud pages via confluence-cli.sh. Uses the same Atlassian credentials as Jira.

Script location: {baseDir}/confluence-cli.sh

Commands

List spaces

bash

{baseDir}/confluence-cli.sh spaces

Returns: [{ id, key, name, type, status }]

Note: most commands use the space id (numeric), not the key.

List pages in a space

bash

{baseDir}/confluence-cli.sh pages 98312
{baseDir}/confluence-cli.sh pages 98312 50

Returns: [{ id, title, status, parentId, authorId, created, url }]

Default limit: 25.

Get page content

bash

{baseDir}/confluence-cli.sh get 12345

Returns: { id, title, status, spaceId, parentId, version, body_text, body_html, truncated, created, url }

body_text is HTML-stripped plain text (first 3000 chars). body_html is raw storage format (first 5000 chars). If truncated is true, the page has more content than shown.

Find page by title

bash

{baseDir}/confluence-cli.sh get-by-title 98312 "Deployment Runbook"

Returns page content directly. Use this when you know the exact page title — avoids a separate search + get.

List child pages

bash

{baseDir}/confluence-cli.sh children 12345

Returns: [{ id, title, status, url }]

Search with CQL

bash

{baseDir}/confluence-cli.sh search "space=ENG AND type=page AND title~\"architecture\""
{baseDir}/confluence-cli.sh search "label=runbook" 20

Common CQL patterns:

• Pages in a space: space=ENG AND type=page
• By title: title~"deployment guide"
• By label: label=runbook
• Recently modified: lastModified > now("-7d")
• By creator: creator=currentUser() AND type=page

Returns: { total, results: [{ id, title, type, space, url }] }

Create page

bash

{baseDir}/confluence-cli.sh create --space 98312 --title "Deployment Runbook" --parent 12345 --body "<h2>Steps</h2><p>1. Pull latest main...</p>"

Required: --space, --title. Optional: --parent (page ID to nest under), --body (HTML storage format).

Returns: { id, title, url }

Update page

bash

{baseDir}/confluence-cli.sh update 12345 --title "Updated Title" --body "<p>New content</p>"

Auto-increments the version number.

Returns: { id, title, version, url }

Get page comments

bash

{baseDir}/confluence-cli.sh comments 12345

Returns: { count, comments: [{ id, body, created, version }] }

Add comment to page

bash

{baseDir}/confluence-cli.sh add-comment 12345 "This section needs updating after the API migration"

Returns: { id, pageId, created }

List attachments

bash

{baseDir}/confluence-cli.sh attachments 12345

Returns: [{ id, title, mediaType, fileSize, downloadUrl }]

Get page labels

bash

{baseDir}/confluence-cli.sh labels 12345

Returns: [{ name, prefix }]

Add labels

bash

{baseDir}/confluence-cli.sh add-labels 12345 "runbook,production,v2"

Labels are comma-separated.

Exploration Rules

When working with Confluence:

1. Search first — use search with CQL to find pages. Do not list all pages in a space and scan through them.
2. Use get-by-title when you know the page name — it's one call instead of search + get.
3. Check truncated — if a page is truncated, tell the user the content was too long to load fully.
4. Use children to navigate page hierarchies instead of listing the whole space.
5. Before creating, always search to check if a similar page already exists.

Body Format

Confluence uses HTML "storage format":

• <p>text</p> — paragraph
• <h2>text</h2> — heading
• <table><tbody><tr><th>Header</th></tr><tr><td>Cell</td></tr></tbody></table> — table
• <strong>text</strong> — bold
• <a href="url">text</a> — link
• <ac:structured-macro ac:name="jira"><ac:parameter ac:name="key">PROJ-123</ac:parameter></ac:structured-macro> — Jira link macro

Rules

• All output is JSON to stdout, errors to both stderr and stdout.
• Never delete pages without explicit user confirmation.
• Always search before creating to avoid duplicates.
• Results are paginated: pages (max 25), spaces (max 25), comments (max 25). Tell the user if results may be incomplete.