Enables access to LLMs via remote providers (OpenAI, Claude, Gemini) and local LLMs via GPT4ALL.
The Large Language Models (LLMs) Plugin gives Obsidian users access to LLMs through cloud providers (OpenAI, Anthropic, Google, and Mistral) and locally via GPT4All and Ollama. Models can be interacted with in the sidebar, main window, and a newly added, floating action button popup window.
Installation: Download the plugin via the community plugin browser.
Using models from cloud-based providers:
Using models locally (GPT4All):
Using models locally (Ollama):
http://localhost:11434)| Command | Description |
|---|---|
| Open modal | Opens the chat modal |
| Toggle FAB | Toggles the visibility of the Floating Action Button (FAB) used to open and close the chat popup window |
| Open chat in tab | Opens the chat window in a tab |
| Open chat in sidebar | Opens the chat window in the Sidebar |
Cloud-based:
| Model provider | Status |
|---|---|
| OpenAI | Supported |
| Anthropic | Supported |
| Supported | |
| Mistral | Supported |
Local:
| Model provider | Status |
|---|---|
| GPT4All | Supported |
| Ollama | Supported |
The plugin can run a built-in MCP (Model Context Protocol) server so Claude Desktop, or any other MCP-compatible client, can read and write this vault directly — no separate companion plugin required.
Setup:
claude_desktop_config.json snippet.{
"mcpServers": {
"obsidian-vault": {
"url": "http://127.0.0.1:27125/mcp",
"headers": {
"Authorization": "Bearer <your-token>"
}
}
}
}
Notes:
127.0.0.1 only and requires the bearer token on every request — it is not reachable from other devices on your network.list_files, read_file, search_vault (read-only, run immediately) and create_file, edit_file, move_file, delete_file (each pops a confirmation dialog in Obsidian before running, the same as in-app agent actions).