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., "@docmost-mcp search the wiki for the onboarding checklist"
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.
An MCP server for self-hosted Docmost. Read, search, create and update wiki pages in Markdown from Claude Code, Claude Desktop, Cursor or any other MCP client.
Docmost's official MCP server and API keys are part of its paid Enterprise edition. This server needs neither. It logs in with a normal Docmost account and calls the same API the Docmost web app uses. Since Docmost v0.70 that API accepts and returns Markdown, and the server converts it to and from its editor format.
Tool | What it does |
| Lists the spaces the account can see, with their ids |
| Lists the root pages of a space, or the children of a page |
| Full-text search, optionally within one space |
| Returns a page as Markdown, with its id, space, parent and URL |
| Creates a page from Markdown, at a space root or under a parent page |
| Replaces, appends to or prepends to a page body; can rename it or change its icon |
Every tool that takes a page accepts a page id, a slug id, or the page URL copied from the browser.
Content updates go through Docmost's real-time collaboration layer, so if someone has the page open while it is updated, their editor picks up the change instead of overwriting it.
Related MCP server: Docmost Community Bridge
Node.js 20.12 or newer
Docmost v0.70 or newer
A Docmost account without MFA. Edits are attributed to this account in page history, so each person should use their own.
There is nothing to clone. Register the server with your MCP client and pass your Docmost login as environment variables. npx downloads and runs it.
claude mcp add docmost --scope user \
-e DOCMOST_URL=https://docs.example.com \
-e DOCMOST_EMAIL=you@example.com \
-e DOCMOST_PASSWORD='your-password' \
-- npx -y @punyamsingh/docmost-mcpAdd the following to the client's MCP config (claude_desktop_config.json, ~/.cursor/mcp.json, …):
{
"mcpServers": {
"docmost": {
"command": "npx",
"args": ["-y", "@punyamsingh/docmost-mcp"],
"env": {
"DOCMOST_URL": "https://docs.example.com",
"DOCMOST_EMAIL": "you@example.com",
"DOCMOST_PASSWORD": "your-password"
}
}
}
}To run the latest code on GitHub instead of the npm release, replace @punyamsingh/docmost-mcp with github:punyamsingh/docmost-mcp. The first start is slower because it builds from source.
The client stores these values in its own config file in plain text, the same as a .env file would, so use a config file only you can read.
After the first login the session token is cached in ~/.cache/docmost-mcp/ (or $XDG_CACHE_HOME/docmost-mcp/), with file mode 600. Restarting the client therefore does not open a new Docmost session each time. A new login happens automatically when the token expires or is revoked.
git clone https://github.com/punyamsingh/docmost-mcp.git
cd docmost-mcp
npm install # also builds into build/
cp .env.example .env # then fill in DOCMOST_URL, DOCMOST_EMAIL, DOCMOST_PASSWORD
claude mcp add docmost --scope user -- node "$PWD/build/index.js".env is git-ignored. Variables set in the client's config take precedence over it.
Uploading local image files is not supported yet; reference images that are already hosted somewhere by URL.
No tools for deleting or moving pages, comments, or permissions.
Accounts with MFA cannot log in, because the server has no way to complete the second factor.
MIT
Creates a page from Markdown at the root of a space or under a parent page, and returns its URL.
Inputs: title (required), markdown (required), icon, spaceId, +1 more.
Returns a page as Markdown along with its id, space, parent and URL.
Inputs: pageId (required).
Lists the top-level pages of a space, or the direct children of a page when parentPageId is given.
Inputs: spaceId, parentPageId.
Lists the Docmost spaces this account can see, with the space ids the other tools take.
Inputs: no named parameters.
Full-text searches page titles and bodies, optionally within one space.
Inputs: query (required), limit, spaceId.
Replaces, appends to or prepends to a page body in Markdown, and can rename it or change its icon.
Inputs: pageId (required), icon, title, markdown, +1 more.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 6 tools
Each tool targets a distinct operation: browsing spaces, browsing page hierarchy, full-text search, single-page fetch, page creation, and page editing. list_pages (hierarchical browsing) and search_pages (full-text lookup) have clearly separate intents, and get_page vs list_pages vs search_pages are easy to tell apart by their outputs.
All six tools follow a clean verb_noun snake_case pattern (list_spaces, list_pages, search_pages, get_page, create_page, update_page). Verbs are used consistently for their action type and no naming style is mixed.
Six tools is well-scoped for a wiki/docs service, covering the natural read/browse/search/create/edit verbs without redundancy. Each tool clearly earns its place in the set.
Create, read (get/list/search), and update are all covered with Markdown fidelity, which handles the core lifecycle. However, there is no delete or archive operation and no explicit page-move/reparent tool, leaving a minor lifecycle gap agents cannot work around.
Shared project wiki for AI agents: read and write pages, next actions, and activity logs over MCP.
Hosted markdown project wikis your team's AI assistants read, search, and update over MCP.
Read and write your team's shared, AI-readable wiki from any MCP client.
Read-only MCP for the Eco game wiki: search, Markdown pages, and wiki_* lookups. No keys, no writes.
