Confluence Integration
The Confluence integration lets your company connect Atlassian Confluence Cloud to IntuitivePM. Use it when Confluence pages should become trusted context for agent workflows without giving agents Atlassian credentials.
With Confluence connected, company admins can:
- Connect Confluence through Atlassian OAuth.
- Discover Confluence spaces.
- Enable only the spaces IntuitivePM may use.
- Search and preview pages from enabled spaces.
- Let approved agents read selected Confluence pages through IntuitivePM.
- Review audit events that show which page an agent read.
Confluence is an Enterprise integration and may need to be enabled for your company before it appears in the Integration Hub.
How It Works
IntuitivePM connects to Confluence through Atlassian OAuth. The Atlassian token is stored and used by IntuitivePM on the server side. Users and agent runtimes never receive Atlassian credentials.
When an agent needs Confluence context, IntuitivePM checks the workflow lease, selected Knowledge Source, enabled space, configured page or page search, and active company. IntuitivePM then reads the selected page through the server-side gateway, converts it to sanitized Markdown, records citations and content hashes, and gives the agent only the scoped context for its current task or workflow run.
Confluence write-back is not enabled yet. Agents can read approved pages, but they cannot edit, delete, comment on, or publish Confluence content.
Prerequisites
- An Atlassian Cloud site with Confluence.
- Company admin access in IntuitivePM.
- Confluence enabled for your IntuitivePM company.
- Permission in Atlassian to authorize the IntuitivePM app.
- Access to the Confluence spaces and pages you want IntuitivePM to use.
- Agent orchestration enabled if you want agents to use Confluence pages as workflow context.
Connect Confluence
- Open System -> Integrations -> Integration Hub.
- Open Confluence.
- Review the connector boundary checklist.
- Select Connect Confluence.
- Sign in to Atlassian when prompted.
- Review and approve the requested permissions.
- Return to IntuitivePM after the OAuth flow completes.
After connection, the Confluence card shows connection status. Open the Confluence details to inspect connection health, connected user, token expiry, enabled spaces, permission drift, and recent operational events.
Enable Spaces
Confluence spaces are disabled by default after connection. This prevents agents from reading a whole Confluence site just because OAuth succeeded.
To enable Confluence content:
- Open System -> Integrations -> Integration Hub.
- Open Confluence.
- Select the Confluence connection.
- Sync spaces.
- Enable only the spaces IntuitivePM may use.
Use the narrowest practical space scope. If only one project needs Confluence context, enable the project space instead of broad company-wide spaces.
Preview Pages
Admins can search and preview pages from enabled spaces before attaching the source to a workflow.
- Open the Confluence details.
- Select an enabled space.
- Enter a page-title search term when needed.
- Select Search pages.
- Select a page to preview the sanitized Markdown, content hash, warnings, citation metadata, and attachment metadata.
Previewing pages confirms that IntuitivePM can read the selected page through the backend gateway. It does not expose Atlassian tokens to the browser or to agents.
Use Confluence As A Knowledge Source
Company admins can create Confluence Knowledge Sources for workflow nodes.
- Open AI Agents -> Knowledge Sources.
- Select New Source.
- Choose Confluence.
- Select the Confluence connection.
- Select the enabled space.
- Select a specific page or configure bounded page discovery for the node.
- Save the source.
- Attach the source to a workflow node that needs that Confluence context.
Confluence Knowledge Sources are most useful for project requirements, technical runbooks, policy pages, implementation notes, customer documentation, and decision records that agents should cite while working.
What Agents Can Read
Agents can receive scoped Confluence page context for the work they are assigned to.
Runtime reads currently include:
- Page title and metadata.
- Sanitized page body as Markdown-like text.
- Citation metadata.
- Content hash.
- Attachment metadata such as file name, MIME type, size, and version.
- Extracted attachment text for supported text, Markdown, CSV, JSON, XML, PDF, DOCX, and XLSX files.
Attachment files are read through IntuitivePM’s backend proxy. Agents do not receive Atlassian download links or credentials. Unsupported binary files, archives, executables, unknown file types, and attachments larger than 10 MiB are blocked by default.
Agents receive page context, source metadata, citations, and content hashes. They do not receive Atlassian credentials and they do not browse Confluence directly.
Audit Source Reads
Company admins can review Confluence source activity from the Confluence connection details and from connector audit diagnostics.
Audit filters can narrow events by:
- Confluence connection.
- Space key.
- Page ID.
- Action.
- Actor type.
- Source type.
- Time window.
This is useful when you need to answer questions such as:
- Did an agent read a Confluence page during a workflow run?
- Which enabled space did the read come from?
- Which page was read?
- Which attachment was read?
- Which agent, runtime, task, or workflow run was involved?
- How much content was returned?
Detailed connector read-audit rows are kept for 180 days by default.
Security Model
IntuitivePM uses these guardrails for Confluence:
- Atlassian credentials stay on the IntuitivePM server.
- Agent runtimes never receive OAuth tokens.
- Confluence reads are scoped to the active company, connection, workflow, node, and lease.
- Spaces are disabled until an admin enables them.
- Source content is treated as untrusted data and sanitized before agents use it.
- Runtime page and supported attachment reads return citations and content hashes.
- Runtime attachment reads are capped at 10 MiB and remain behind company rollout controls.
- Confluence write-back is not enabled in the current rollout.
- Confluence reads and operational events are auditable.
Troubleshooting
| Issue | What to check |
|---|---|
| Confluence card is missing | Confirm Confluence is enabled for your company and that you have admin access. |
| OAuth connection fails | Confirm the Atlassian account can approve the requested Confluence permissions. |
| No spaces appear | Sync spaces and confirm the connected Atlassian user can access the expected Confluence site. |
| Space cannot be used by agents | Confirm the space is enabled in the Confluence connection details. |
| Page preview returns no results | Confirm the space is enabled and try a broader page title search term. |
| Agent cannot read a Confluence page | Confirm the workflow node uses a Confluence Knowledge Source that points to an enabled space and valid page. |
| Attachment content is missing | Confirm the attachment is on the selected page, uses a supported file type, and is below 10 MiB. Unsupported binaries still need to be summarized on the page or moved into an approved Work File. |
| Connection requires re-authentication | Open the Confluence card and re-authenticate the connection. |
Recommended Setup
For most teams:
- Connect Confluence from the Integration Hub.
- Enable only one pilot space.
- Preview one or two pages from that space.
- Create a Confluence Knowledge Source from a specific page.
- Attach it to a low-risk workflow node.
- Run a test task and confirm the answer includes citations.
- Review audit events before enabling more spaces.