Folio

What Folio MCP server does

Folio MCP server connects AI agents to content managed by Folio, a self-hosted team wiki built around a Git repository. The available capabilities cover searching the wiki, reading pages, and making changes to pages, data tables, and whiteboards. This lets an agent use the same content that people edit in Folio’s browser interface.

Folio is intended for documentation that benefits from repository ownership and history. Pages are stored as files in a Git repository rather than existing only in an application database. Markdown pages, table files, whiteboards, attached files, and their surrounding folder structure can therefore remain available outside Folio as well.

How it works

Folio keeps the repository as the authoritative home for wiki content. Its database maintains supporting information such as the search index, access permissions, and live-editing state. Browser edits and agent changes operate on the same pages, so an agent can read an existing document and update it without creating a separate copy of the wiki.

Git provides the content history. Changes can be reviewed through commits, diffs, blame, merge requests, and reverts. The repository layout also supports ordinary relative Markdown links, documentation folders, and formats used by tools such as Docusaurus, MkDocs, Hugo, and Jekyll. These repository behaviors may matter when an agent edits documentation that is also maintained through developer workflows.

Folio supports several content types. A data table is represented by a Markdown file with front matter for table metadata and a readable table for its rows. A whiteboard is stored as an SVG that retains its editable scene. The README also describes real-time editing, access rights, page history, and offline work for supported browser interactions.

Setup and configuration

The repository can be started with Docker Compose. The documented first-run sequence clones the repository, changes into its directory, and runs docker compose up -d; the web application is then opened at http://localhost:4870 to create the first account. This is the documented setup for Folio itself.

The supplied material does not specify a separate MCP installation command, transport, authentication setting, endpoint, or environment variable. It also does not identify whether the MCP interface requires additional configuration beyond running the Folio instance. Developers should verify those details in the project’s current documentation before wiring an agent to a deployment.

Tools and capabilities

The Folio MCP server is described as a full MCP server for the wiki. Its documented agent-facing scope includes:

• Searching Folio content.
• Reading wiki pages.
• Editing pages.
• Working with data tables.
• Working with whiteboards.

The README’s example workflow has an agent read a release process, write release notes, and add a row to a roadmap. That demonstrates page and table updates, but the supplied material does not provide formal tool names, schemas, argument definitions, or error behavior.

The system is suited to teams that want agents to operate on shared documentation rather than on an isolated knowledge export. Because the underlying content remains in Git, repository-level review and recovery practices can remain part of the documentation workflow.

Limitations and notes

The supplied README excerpt does not list supported MCP clients such as Cursor or Claude Desktop. It also does not document the MCP transport, permission model for agent calls, setup-specific environment variables, or whether every Folio content type supports offline agent access. Browser offline support is explicitly limited: pages and whiteboards can be edited without a connection, while data tables, forms, and file uploads require one.

The public demo is shared, resets its data every 24 hours, and advises users not to enter personal data or API keys. A self-hosted deployment is the documented option for maintaining a team’s own content.