Skip to main content
Glama

@fswap/mcp-outline

An MCP server for Outline that lets Claude Desktop, Claude Code, Cursor and other MCP clients search, read, create and update your wiki.

Runs locally over stdio. No install step — clients launch it with npx.

Quick start

  1. Create an API token in Outline under Settings → API.

  2. Run the interactive setup once:

    npx -y @fswap/mcp-outline@latest setup

    It asks for your Outline URL and token, verifies them against auth.info, and stores them in your OS config directory (mode 0600).

  3. Add the server to your client. At the end, setup offers to do this for you in Claude Code, Claude Desktop, Cursor and Codex: clients it finds are pre-selected (except Claude Code when Claude Desktop is installed, because the desktop app's Code tab already loads claude_desktop_config.json), an existing outline entry is only replaced after you confirm, and any config file it changes gets a one-time .bak copy.

    To add it by hand instead, use the snippets below. Every value asked in setup can be skipped with Enter; anything you skip goes into the env block instead. setup --print shows these snippets again at any time.

    Claude Desktop (claude_desktop_config.json) and Cursor (~/.cursor/mcp.json or <project>/.cursor/mcp.json):

    {
      "mcpServers": {
        "outline": {
          "command": "npx",
          "args": ["-y", "@fswap/mcp-outline@latest"]
        }
      }
    }

    Codex (~/.codex/config.toml):

    [mcp_servers.outline]
    command = "npx"
    args = ["-y", "@fswap/mcp-outline@latest"]

    Claude Code:

    claude mcp add outline -- npx -y @fswap/mcp-outline@latest

Without setup (environment variables)

Environment variables take precedence over the config file, so you can skip setup entirely (or skip individual values in it and set them here):

{
  "mcpServers": {
    "outline": {
      "command": "npx",
      "args": ["-y", "@fswap/mcp-outline@latest"],
      "env": {
        "OUTLINE_URL": "https://app.getoutline.com",
        "OUTLINE_API_TOKEN": "ol_api_..."
      }
    }
  }
}

Codex equivalent:

[mcp_servers.outline]
command = "npx"
args = ["-y", "@fswap/mcp-outline@latest"]
[mcp_servers.outline.env]
OUTLINE_URL = "https://app.getoutline.com"
OUTLINE_API_TOKEN = "ol_api_..."

Variable

Purpose

OUTLINE_URL

Outline base URL (cloud or self-hosted)

OUTLINE_API_TOKEN

API token

OUTLINE_ALLOW_DELETE

true to expose delete_document

OUTLINE_DEFAULT_COLLECTION

Optional default collection name

Related MCP server: outline-mcp

Tools

Tool

Outline endpoint

Notes

list_collections

collections.list

id, name, description, url

get_collection

collections.info

includes document tree

search_documents

documents.search

query, optional collectionId; returns snippets, not bodies

get_document

documents.info

full markdown body; accepts id or URL slug

list_documents

documents.list

filter by collectionId / parentDocumentId

create_document

documents.create

title, markdown, collection, optional parent; published by default

update_document

documents.update

title/text; append=true appends instead of replacing

move_document

documents.move

change collection and/or parent

archive_document

documents.archive

reversible

delete_document

documents.delete

only when delete is allowed (setup answer or OUTLINE_ALLOW_DELETE=true)

API errors are returned to the model as isError results rather than crashing the server.

Development

TypeScript source in src/, bundled to dist/ with tsdown. Only dist/ is published.

npm install
npm run build        # tsdown → dist/index.js
npm run lint         # eslint (typescript-eslint)
npm run typecheck    # tsc --noEmit
npm test             # builds, then spawns the server and checks the tool list
npm run check        # all of the above (also runs on prepublishOnly)
OUTLINE_URL=... OUTLINE_API_TOKEN=... npm run inspect   # MCP Inspector UI against dist/

Never write to stdout from server code — it is the protocol channel. Use console.error.

Reset

npx -y @fswap/mcp-outline@latest setup --reset

License

MIT

Available Tools

9 tools
archive_documentArchive documentA

Archive a document (reversible; it disappears from the collection but can be restored from the archive). Prefer this over delete_document.

Inputs: id (required).

create_documentCreate documentA

Create a new Outline document from markdown. Published by default; pass publish=false to save a draft. Requires a collectionId (use list_collections). Optionally nest under a parent document.

Inputs: title (required), collectionId (required), text, publish, +1 more.

get_collectionGet collectionA

Get one Outline collection including its document tree (ids, titles and nesting). Useful to understand how a collection is organised before creating or moving documents.

Inputs: id (required).

get_documentGet documentA

Fetch a single Outline document including its full markdown body. Accepts a document id, a urlId (the slug part of the URL) or a share id.

Inputs: id (required).

list_collectionsList collectionsA

List Outline collections (top-level groupings of documents). Returns id, name, description and url. Use the id with search_documents, list_documents or create_document.

Inputs: limit, query.

list_documentsList documentsA

List documents, optionally filtered by collection and/or parent document. Returns id, title, url and timestamps (no bodies). Sorted by most recently updated by default.

Inputs: sort, limit, offset, direction, +2 more.

move_documentMove documentB

Move a document to another collection and/or under a different parent document. Omit parentDocumentId to move it to the top level of the collection.

Inputs: id (required), index, collectionId, parentDocumentId.

search_documentsSearch documentsA

Full-text search across Outline documents. Returns id, title, url and a short context snippet per hit. Does NOT return document bodies — call get_document with an id to read one.

Inputs: query (required), limit, offset, collectionId.

update_documentUpdate documentA

Update an existing document's title and/or markdown body. By default text REPLACES the body; set append=true to add text to the end instead. Only the fields you pass are changed.

Inputs: id (required), done, text, title, +2 more.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4/5.0

    Scored across 9 tools

    Disambiguation5/5

    Each tool has a clearly distinct action-resource pairing: list/search/get/list-collections for reads, create/update/move/archive for writes. The only mild overlap is list_documents vs search_documents, but descriptions make the boundary explicit (metadata listing vs full-text search with snippets) and search_documents points to get_document for bodies.

    Naming Consistency5/5

    All nine tools follow a strict snake_case verb_noun pattern (list_collections, get_collection, list_documents, create_document, update_document, move_document, archive_document, search_documents, get_document) with no exceptions or mixed conventions.

    Tool Count5/5

    Nine tools is well-scoped for an Outline document service, covering both the collection and document resources without redundancy. Every tool earns its place, none feels like filler.

    Completeness4/5

    The document lifecycle is largely covered (list, search, get, create, update, move, archive), plus collection reads. However, there is no collection creation/update and no explicit unarchive/restore or hard delete, leaving minor gaps agents must work around; archive is documented as reversible but no restore tool is exposed.

    Maintenance

    ActivitySlowing
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers