mediawiki-mcp
mediawiki-mcp connects an AI assistant to one or more MediaWiki sites so it can search, read, edit, organize, and monitor wiki content.
Register, remove, list, and set default wikis, including authentication status.
Locate pages via exact/redirect lookup, title-prefix search, or full-text search across one or all wikis.
Read page wikitext, metadata, rendered HTML optionally, outgoing links, and backlinks.
Create pages or update existing pages using diff, section, append, prepend, or full-content edits.
Delete and undelete pages when the configured account has admin/sysop rights.
Inspect page history, individual revisions, categories, category members, files, and recent changes.
Upload files from base64 data or a source URL, with attribution and comments.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@mediawiki-mcp search for 'MCP' on the Main wiki"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
MediaWiki MCP
Connect an AI assistant to your MediaWiki sites. Ask it to find and summarize pages, compare revisions, or prepare edits. One connection can serve several wikis; writing uses the permissions of the wiki account you configure.
Choose your setup
Start with one guide. You do not need to follow all four.
Where you want to use it | Start here | What you need |
Claude Desktop on your computer | Claude Desktop and a wiki URL | |
Claude Code or another local MCP client | Node.js 24 LTS recommended; minimum Node.js 20 | |
LibreChat or a private HTTP service | Docker Compose and access to the wiki | |
A public connector with Microsoft sign-in | Public HTTPS address, Entra app, and a server |
Your wiki must run MediaWiki 1.42 or later, with its REST and Action APIs
reachable. Use the wiki's API base URL, such as https://wiki.example.com or
https://example.com/w; see how to choose the URL.
Anonymous access works only where the wiki permits it. For private reads or edits,
create a bot password.
Related MCP server: mediawiki-mcp-server
Check your connection
After completing your chosen guide, ask your assistant:
List the connected wikis and their authentication status.
It should call list-wikis and show your wiki. Then try:
Find the page about backups and summarize it. Include the wiki and page title.
The usual workflow is find-page → get-page → a summary or proposed edit.
For editing, read the page first and review the proposed change before saving.
See working with pages for concrete examples.
What you can do
Search one wiki or all configured wikis, then read pages and follow links.
Inspect categories, files, revision history, and recent activity.
Create or update pages and upload files when your account has the required rights. OAuth connections also require the configured write role.
The tool reference lists the exact tool names, inputs, and which ones are available in each connection mode.
Find the right reference
I need to… | Read |
Set wiki URLs, credentials, multiple wikis, or environment variables | |
Understand sign-in, permissions, image restrictions, or session expiry | |
Upgrade, roll back, or diagnose a failed connection | |
Change the code or run a disposable test wiki | |
Prepare or verify a release |
For a full navigation page, open the documentation index. Published versions and Desktop bundles are on GitHub Releases. Changes are recorded in the changelog.
License
MIT.
Available Tools
21 toolsadd-wikiA
Register a new MediaWiki instance so you can read and edit its pages. Provide the wiki base URL and optional bot credentials. After adding, use the wiki name in other tools' "wiki" parameter.
Inputs: url (required), name (required), password, username.
create-pageA
Create a new wiki page with wikitext content. The page must not already exist — use update-page to modify existing pages.
Inputs: title (required), content (required), summary (required), user, +1 more.
delete-pageA
Delete a wiki page permanently. This requires admin/sysop rights on the wiki.
Inputs: title (required), user, wiki, reason.
find-pageA
Unified page locator. Tries exact title (following redirects), then title-prefix match, then full-text search, across all registered wikis by default. Returns a single ranked list where exact/redirect hits come before prefix hits, which come before full-text hits. Use this as the first step whenever you need to locate a specific page — it succeeds regardless of whether the user gave you the exact title, a partial title, or a topic description.
Inputs: query (required), wiki, limit.
get-category-membersA
List all pages, subcategories, or files within a specific category. Returns member titles and IDs. Use type to filter (page/subcat/file).
Inputs: category (required), type, wiki, limit, +1 more.
get-fileB
Get metadata and download URLs for an uploaded file (image, PDF, etc). Returns dimensions, file size, media type, and both original and preferred-format URLs.
Inputs: title (required), wiki.
get-pageA
Retrieve a wiki page's full wikitext source and metadata. Returns the page title, ID, content model, latest revision ID/timestamp, and the raw wikitext source. Call this BEFORE update-page to read the current content you want to edit.
Inputs: title (required), wiki, include_html.
get-page-historyA
Get the revision history of a wiki page. Returns a list of revisions with IDs, timestamps, authors, size changes, and edit summaries. Use older_than to paginate through long histories.
Inputs: title (required), wiki, limit, older_than.
get-page-linksA
Get outgoing links from a page (direction="from") or find all pages that link to it (direction="to", i.e. backlinks). Returns link titles and namespaces.
Inputs: title (required), wiki, limit, direction, +1 more.
get-recent-changesA
Get a feed of recent edits, page creations, and log events across wikis. Returns timestamps, change types, page titles, authors, size deltas, and edit summaries. Useful for monitoring wiki activity.
Inputs: type, wiki, limit, namespace, +1 more.
get-revisionB
Get details for a specific revision by ID. Returns the page title, timestamp, author, size, byte delta, and edit comment.
Inputs: revision_id (required), wiki.
list-categoriesB
List all categories on the wiki with page/subcat/file counts. Use prefix to filter by name prefix. Useful for discovering how content is organized.
Inputs: wiki, limit, prefix, continue_from.
list-wikisA
List all registered wikis with their URLs, auth status, and which is the default. Call this first if you're unsure which wikis are available.
Inputs: no named parameters.
remove-wikiA
Unregister a wiki so it is no longer accessible. Does not affect the wiki itself — only removes it from this session.
Inputs: name (required).
search-pageA
Full-text search across wiki pages. Returns matching page titles, IDs, and text excerpts. Use this to find pages when you don't know the exact title.
Inputs: query (required), wiki, limit.
search-page-by-prefixA
Search for pages by title prefix (autocomplete-style). Returns matching page titles and IDs. Faster than full-text search when you know the beginning of the page title.
Inputs: query (required), wiki, limit.
set-wikiA
Set the default wiki. When multiple wikis are registered, this determines which wiki is used when the "wiki" parameter is omitted from other tool calls.
Inputs: name (required).
undelete-pageA
Restore a previously deleted wiki page. This requires admin/sysop rights on the wiki.
Inputs: title (required), user, wiki, reason.
update-pageA
Update an existing wiki page. WORKFLOW: First call get-page to read the current wikitext source, then call this tool with your changes.
EDIT MODES (use exactly one): • diff: Preferred for targeted edits. Provide a unified diff (with @@ hunk headers and context lines) — the server fetches current content, applies the patch, and saves. Best for surgical changes to large pages. • section + content: Edit a single section by number (0 = lead section). Only that section is replaced. • append: Add text to the end of the page without reading current content first. • prepend: Add text to the beginning of the page without reading current content first. • content: Full page replacement. Avoid for large pages — use diff instead.
WIKITEXT STYLE GUIDE — when writing or updating wikitext, use modern syntax: • Tables: {| class="wikitable" with |- row separators (not |---- or border="1" or HTML ) • Bold/italic: '''bold''' and ''italic'' (not /) • Headings: == Level 2 == through ====== Level 6 ====== (skip level 1) • Lists: * bullets, # numbered, ; and : for definition lists • Links: [URL description] for external (not bare URLs) • Avoid deprecated tags: , , , , , • Avoid deep colon indentation (::), excessive , inline CSS on divs
Inputs: title (required), summary (required), diff, user, +5 more.
upload-fileC
Upload a file to the wiki from base64-encoded data. Provide the file content as a base64 string.
Inputs: data (required), filename (required), description (required), user, +2 more.
upload-file-from-urlA
Upload a file to the wiki by fetching it from a URL. The wiki server downloads the file directly from the source URL.
Inputs: url (required), filename (required), description (required), user, +2 more.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 21 tools
Most tools target distinct resources, but find-page is explicitly a 'unified page locator' that performs exact, prefix, and full-text search, directly overlapping with search-page and search-page-by-prefix. The descriptions do explain the tradeoffs (speed, unified ranking), so selection is recoverable, but the redundancy is real.
All 21 tools use consistent kebab-case with a predictable verb-noun pattern (get-page, create-page, delete-page, list-wikis, add-wiki), with only natural multi-noun extensions like get-page-history and get-category-members. No mixing of conventions.
21 tools is on the heavy side for the 3-15 sweet spot, but each addresses a distinct MediaWiki operation: page CRUD, history, categories, files, search, and wiki registration. Nothing appears redundant enough to cut except possibly the overlapping search trio.
Strong lifecycle coverage: create/read/update/delete/undelete pages, revision history, revisions, links/backlinks, categories and members, file metadata and two upload paths, recent changes, and multi-wiki management. Minor gaps like page move/rename and protection changes exist but are workaroundable via wikitext edits.
Maintenance
Related MCP Connectors
One MCP server for all your MCP servers: code mode, per-tool permissions, approvals and audit.
Unofficial Wikipedia MCP — wraps Wikipedia REST API (free, no auth)
An MCP server that provides read access to your cloud storage providers, bank accounts and more.
Related MCP Servers
- AlicenseBqualityBmaintenanceA secure MCP server for interacting with MediaWiki instances, allowing users to search, read, create, and manage wiki content like pages, categories, and files. It supports both public and private wikis with comprehensive authentication for full read and write operations.19AGPL 3.0
- AlicenseAqualityAmaintenanceMCP server for MediaWiki wikis. Search, read, edit, and manage wiki content from AI assistants. Includes formatting, link checking, revision history, and markdown conversion.4325MIT
- AlicenseNot gradedqualityBmaintenanceMCP server for Wiki.js with full GraphQL API coverage, fine-grained permissions, multi-user support, and deployable on Vercel.Apache 2.0
- AlicenseNot gradedqualityDmaintenanceMCP server providing live Wikipedia recent changes feed, page summaries, trending pages, and Wikidata entity lookup.MIT