Can AI agents automate Obsidian?

Obsidian is no longer accurately described as having no official automation interface. Vault content remains ordinary local Markdown and attachments, so agents can safely read, diff, generate, and version notes with standard file tools while leaving application settings in the per-vault `.obsidian` directory. Obsidian also publishes a typed TypeScript plugin API for vault, metadata, editor, command, workspace, and lifecycle events, plus URI actions for opening, creating, appending, searching, and callback workflows. Since version 1.12, the official desktop CLI can read and create notes, search, query Bases, manage properties and tasks, execute application or plugin commands, inspect Sync history, and publish changes with JSON, CSV, TSV, Markdown, or path-oriented output. The desktop application normally must be running. A separate open-beta `obsidian-headless` client can synchronize or publish paid service vaults on servers without the desktop app, including pull-only and mirror modes. Core personal and organizational use is free; Sync, Publish, Catalyst, and optional Commercial support are paid. Plugins execute locally with broad vault and potentially Node or Electron access, so untrusted plugins and direct file edits need backups and an isolated test vault. A first-party MCP server, general external webhook boundary, and repeatable computer-use result were not established.

Best path today: cli · Overall automatability: 8/10