Obsidian Copilot V4: workflows and multi-agent review

Explore Copilot V4 in Obsidian through dashboards, a 7,000-note reorganization, and paid multi-agent reviews with Claude Code, Codex, and OpenCode.

Player not loading? Watch on YouTube

Obsidian Copilot V4 brings AI agent workflows into an Obsidian vault. Logan, a member of the Copilot team, demonstrates a company dashboard built with Dataview, a custom frosted glass theme, and a reorganization of more than 7,000 notes. He says the initial reorganization took less than an hour. His personal filing framework keeps his own writing separate from an agent-maintained wiki so he can trace material back to its source.

The contract-review example puts Claude Code, Codex, and OpenCode in the same conversation. Copilot summarizes their responses, including agreements and disagreements, and lets the user inspect each response. Agents added in later turns can read earlier replies from the other agents. The presenter identifies this multi-agent capability as a paid feature. Custom teams with separate agent memories are described as planned additions.

OpenArtifacts publishes Markdown notes as web pages while the source remains in the vault. Other examples include a custom skill for breaking down YouTube footage into shots and a snake-game plugin created inside Obsidian. Inline comments and private sharing are future plans.

The presenter describes using existing Claude Code or Codex subscriptions, personal keys through OpenCode, or hosted open models. Notes, chats and skills remain in the vault while the user chooses which agent handles each task.