Copilot for Obsidian

★ 7.398

Chat assistant right inside the vault: ask questions about your own notes, draft text — with selectable models, including Claude Opus 4.7.

Copilot for Obsidian is a chat assistant that lives right inside the vault: you mark notes with @ to pull them into the chat as context, and discuss or revise them in a dialogue. The free version is bring-your-own-key: you enter an API key from any provider (OpenRouter, Gemini, OpenAI, Anthropic, Cohere — OpenRouter is recommended) under Settings → Copilot → Basic, after which everything runs directly against the chosen provider, without data passing through Brevilabs servers. It includes a chat mode with note references, a vault-QA mode that searches the entire note collection (embeddings are optional, Copilot delivers instant results even without them), and "Relevant Notes" — suggestions based on semantic similarity and linking. The paid Plus/Believer tier (licensed by Brevilabs LLC, not affiliated with Obsidian or Anthropic) adds an autonomous "Agent Mode" with independent tool-calling, understanding of images/PDFs/EPUBs/YouTube videos/web pages, as well as a "Project Mode" for folder-/tag-based AI context (similar to NotebookLM, but inside your own vault). Brand new is Copilot v4: it lets you run a full-fledged coding agent — opencode, Claude Code, or Codex — natively inside the vault, one that can plan, search the vault, and act, without ever leaving the notes. The frontend code is fully open source, while the backend behind the Plus agent features is proprietary.

🔧 SETUP

In Obsidian: Settings → Community plugins → turn off Safe mode (if active).
Click "Browse", search for "Copilot for Obsidian", click "Install" and then "Enable".
For the free version: Settings → Copilot → Basic → "Set Keys", choose the desired provider (e.g. Anthropic or OpenRouter) and paste in the API key.
For Plus/Believer: copy the license key from the dashboard (obsidiancopilot.com/en/dashboard) and paste it into the "Copilot Plus" card under Settings → Copilot → Basic.
Use `@` in the chat input to add specific notes as context, or use vault-QA mode for questions across the entire vault.

✅ WHEN TO USE IT?

Quick questions to your own notes, no setup

In vault-QA mode you can immediately ask "What did I note about topic X last year?" without having to build embeddings first.

Using a coding agent right inside your second brain

With Copilot v4 you can launch Claude Code natively inside the vault to, for example, automatically build a structured summary from scattered notes, without first copying the notes into a separate project.

⛔ WHEN NOT TO?

Full data control without any cloud provider

Even in the free tier, the chat runs against an external LLM provider (OpenAI, Anthropic, Google …) — completely local inference with no cloud calls at all isn't provided for in the standard setup.

Better: For purely local, offline-capable semantic search with no API calls at all, switch to Smart Connections (local embedding model).

Agent features with full backend transparency

The Plus tier's Agent Mode and file conversion run through proprietary, closed-source Brevilabs servers — only the frontend is open source.

Better: For fully open-source vault automation, use Local REST API with your own MCP client/script.