Workflowy Local MCP
Local Workflowy search and editing for AI assistants.
The app lets MCP clients read, edit, and search Workflowy. Search runs against a local SQLite cache; reads and edits use Workflowy’s APIs.
- 16 tools for reading, editing, searching, and managing Workflowy
- Fast full-text search across locally cached nodes
- Multiple accounts, bookmarks, and custom AI instructions
- Local daily backups with configurable retention
- A desktop interface for setup, configuration, and logs
- Releases for macOS, Windows, and Linux
From server to desktop app
Workflowy’s public API can read and update items, but it cannot search, and full exports are limited to one request per minute. A simple API wrapper would be too slow on a large account, so I added a local full-text index.
The first version worked as a server on my machine. Once other people used it, setup and recovery mattered as much as the tools. Their feedback led to desktop setup, account switching, backups, logs, clearer errors, and installers for macOS, Windows, and Linux.
After the first release, the larger job was making the server understandable and dependable for someone who had not built it.