Guides
MCP server
SlimWiki runs an MCP server, so an AI assistant that supports the Model Context Protocol can read, search and create pages as tools, with nothing to install.
Connect
The server speaks MCP over HTTP (the Streamable HTTP transport) at one address:
https://api.slimwiki.com/mcpThere are two ways to connect. Both act as you, with your access.
- Sign in. Give your assistant the URL. It opens SlimWiki in your browser, you sign in and choose Allow, and it is connected. Nothing to copy.
- API key. For scripts and for clients that cannot open a browser, send a key in a header, as for the API.
Connect by signing in
In an assistant that supports remote MCP servers, add a server (often called a connector) with the URL above. In Claude Code:
claude mcp add --transport http slimwiki https://api.slimwiki.com/mcpThe first time the server is used, your browser opens SlimWiki. The page shows the name the app gave itself, what it will be able to do, and the address it will return you to. Choose Allow to connect or Deny to refuse.
| Permission | Lets the app |
|---|---|
wiki:read | Read the wikis and pages you can see. |
wiki:write | Create, archive and restore pages as you. |
offline_access | Stay connected without asking you to sign in again each hour. |
Connect with an API key
Create a key in SlimWiki under Profile → API keys (see Authentication) and send it as a bearer token. In Claude Code, with the key in the SLIMWIKI_API_KEY environment variable:
claude mcp add --transport http slimwiki https://api.slimwiki.com/mcp \
--header "Authorization: Bearer $SLIMWIKI_API_KEY"Clients configured with a JSON file use an entry of this shape:
{
"mcpServers": {
"slimwiki": {
"type": "http",
"url": "https://api.slimwiki.com/mcp",
"headers": {
"Authorization": "Bearer swk_your_key"
}
}
}
}Tools
| Tool | What it does | Needs |
|---|---|---|
whoami | Returns the user the connection acts as. | Read access |
list_accounts | Lists the accounts the user belongs to. | Read access |
list_wikis | Lists the wikis in an account. | Read access |
list_pages | Lists the pages in a wiki, without their content. | Read access |
search_pages | Searches the titles and content of the pages in a wiki. | Read access |
get_page | Returns one page with its content. | Read access |
create_page | Creates a page, optionally under another page. | Write access |
archive_page | Moves a page to the archive. | Write access |
restore_page | Brings an archived page back. | Write access |
Each tool is one call to the API and returns that endpoint's JSON response, so the API reference describes the fields. A request the API refuses comes back as a tool error carrying the usual error body.
What to expect
- Your access, no more. The assistant acts as the person who signed in or created the key and can reach only what they can. A read-only key, or a connection allowed for reading only, cannot create, archive or restore pages.
- No permanent delete. There is no tool to delete a page. An assistant can archive a page, which can be undone.
- No editing existing pages. As with the API, content is set when a page is created; see Content format.
- Rate limit. Every tool call counts as one request towards the connection's limit of 120 requests per minute. See Rate limits.
- No sessions. The server keeps no state between requests and accepts
POSTonly.