Confluence Workflows

Confluence Workflows

Typical agent tasks against Confluence Data Center. Typed tools cover the common content lifecycle; everything else goes through the generic tools (atlassian_discover_operations → atlassian_describe_operation → atlassian_execute_operation).

Search (CQL)

confluence_search({ cql: "space = MCP AND type = page AND text ~ \"release notes\"" })

Results paginate with cursor / nextCursor; responseProfile and fields control how much each item carries.

Read a page

confluence_get_content({ id: "123456", expand: ["body.storage", "version"] })

Use expand to pull the storage-format body, version, ancestors, or labels in one call.

Create a page (safe tier)

confluence_create_content({
  content: {
    type: "page",
    title: "Release notes 1.0",
    space: { key: "MCP" },
    body: { storage: { value: "<p>Hello <strong>Confluence</strong></p>", representation: "storage" } }
  }
})

The body must be Confluence storage-format XHTML. The server does not convert Markdown — pre-convert with an external tool such as pandoc.

Large pages: storageValueFile

Inline bodies become unwieldy for big pages. Both confluence_create_content and confluence_update_content accept an optional storageValueFile parameter instead of an inline body.storage.value:

confluence_create_content({
  content: {
    type: "page",
    title: "Big design doc",
    space: { key: "MCP" }
  },
  storageValueFile: "/abs/path/under/file-root/design-doc.xhtml"
})

The file (storage XHTML, size-capped — see Attachments and Large Results) is read from the ATLASSIAN_FILE_ROOT sandbox and becomes body.storage.value; representation is fixed to storage. storageValueFile is mutually exclusive with an inline body.storage.value — passing both is a structured error.

Update a page (safe tier)

Updates need the current version number — read first, then update with version.number incremented:

confluence_update_content({
  id: "123456",
  content: {
    type: "page",
    title: "Release notes 1.0",
    version: { number: 2 },
    body: { storage: { value: "<p>Updated</p>", representation: "storage" } }
  }
})

A version conflict means someone else edited the page — re-read and retry deliberately; the server never retries any request automatically.

Attachments

Upload (multipart, safe tier, file root required):

atlassian_execute_operation({
  operationId: "confluence.attachments.upload",
  pathParams: { id: "123456" },
  body: { files: ["/abs/path/under/file-root/diagram.png"] }
})

Download with confluence_download_attachment.

Delete (risky tier)

confluence_delete_content({ id: "123456" })

Deletes are risky — they are not visible unless the server runs with --exposure-tier=risky or higher.

Labels and restrictions (via generic execute)

atlassian_execute_operation({ operationId: "confluence.content.labels.add", pathParams: { id: "123456" }, body: [{ prefix: "global", name: "release" }] })
atlassian_execute_operation({ operationId: "confluence.content.labels.list", pathParams: { id: "123456" } })

Discover the exact operation IDs and parameters with atlassian_discover_operations({ query: "label" }) and atlassian_describe_operation.

---

中文版 · English is the authoritative version.