Krosoft Wiki
Auto Light Dark
Auto Light Dark

Connect your MCP client

Use your AI assistant or IDE to search, read, create and update Confluence pages through MCP. You connect with your own Confluence account and retain its permissions.

Connect your client

  1. In your MCP client, add a connection to MCP for Confluence Data Center using the settings in the table below.

  2. In Confluence, open your avatar → Settings → Personal access tokens and create a token.

  3. If your client provides a Bearer token field, paste only your PAT there. If it uses custom headers, add a header named Authorization with the value Bearer <your-PAT>, replacing <your-PAT> with your token. Use the client’s secret storage or masked credential prompt where available.

  4. Connect and ask your assistant which Confluence account it is using. Confirm it is your account.

For client-specific configuration and secure PAT entry, see Set up VS Code (GitHub Copilot), Set up Cursor, or Set up Microsoft Copilot Studio.

Setting

Value

Transport

Streamable HTTP

URL

<confluence-base-url>/rest/native-mcp/1.0/mcp

Header

Authorization: Bearer <your-Confluence-PAT>

Replace <confluence-base-url> with your Confluence site’s address, including any context path such as /confluence. Keep certificate verification enabled. Keep your token out of prompts, shared files and source control.

Ask in plain language

You do not need to know query syntax or page IDs. Your assistant prepares the tool calls. For example:

  • “Find the architecture decisions about database migrations and cite the pages.”

  • “Find our onboarding guide in the Engineering space.”

  • “Read this page and summarize the decision, with a link to the source.”

Results depend on your Confluence access and any optional administrator space filter. Search uses Confluence’s index, so recent content may not appear immediately. A partial page read or search batch is not a complete inventory.

Create or update a page

Ask your assistant to draft a page or propose a specific change. Review the intended location and content before asking it to save. Your account needs Confluence permission to create or edit the target page.

Full-page updates replace the complete page body. A section change can instead preserve the rest of a supported page. If the assistant cannot read the full page, use the Confluence editor. Ask it to preserve unrelated content and show the saved result. The technical reference explains the editing limits.

Get help

If the connection or access fails, see Troubleshooting. For authentication and access rules, see Security and permissions. Developers can use Developer and technical integration for the protocol and tool contract.

Ask in plain language, for example:

  • “Add this comment to the decision page, and reply to the earlier discussion.”

  • “Label this page architecture-decision without removing its other labels.”

  • “Show the previous version and who changed it.”

  • “Attach this small file, then copy it to the other page if I have access.”

  • “Change only the second Decision section and keep the rest of the page intact.”

Comments follow your native author, edit and delete permissions. Small attachments are limited to 32 KiB; an existing filename is rejected, so choose another name. Section changes require a precise target and a fresh page version. If headings are duplicated or a change conflicts, ask the assistant to identify the exact section again. Pages too large to read completely, and headings nested inside complex containers, require the Confluence editor.

After a timeout or uncertain save, inspect the native page, comment or attachment before asking the assistant to repeat it. The change may already have been saved.

Browse, reuse and organize documentation

Use these workflows with your existing Confluence account. Native permissions and administrator policies apply to every operation.

  • “Find my project space and show where this page belongs.” Browse spaces, page ancestry and small batches of children. Hidden pages and spaces remain hidden.

  • “Upload this file, then replace it with this new version.” Larger files transfer in small pieces, with a default maximum of 10 MiB. A new version is an explicit choice; an ordinary upload still refuses an existing filename.

  • “Copy this documentation tree and include its files and labels.” Copies can include up to 25 pages and five levels. Restricted source pages require native Confluence. Copies inherit destination access, and links/macros keep their original targets: review references before reusing the documents.

  • “Move this page under the project overview, before the checklist.” Moves and sibling ordering work within one space. The destination can change who inherits access. Use native Confluence for moves between spaces.

  • “Trash this obsolete page, then restore it under the project overview.” Only pages without children can be trashed through the assistant. Trash discovery and restoration require the space administrator. Permanent deletion is not offered.

After an uncertain or partially completed change, ask the assistant to check the native result before continuing. Previously copied pages may remain after a later item fails; repeating the whole operation can create duplicates.