OpenAgent

An AI agent that lives inside your vault — vault-aware, tool-capable, BYOK, and cross-platform (desktop + mobile).

Hackathon build: See hackathon/README.md for the Gemma 4 Good submission story, local MLX setup, eval results, and demo assets.

Features

• OpenAI-compatible provider — works with OpenAI, Anthropic (via proxy), Ollama, LM Studio, or any OpenAI-compatible endpoint
• Vault tools — read, write, edit, append, search, list notes, manage frontmatter and links
• Consent & safety — per-action confirmation dialogs with diff previews before any destructive write
• BYOK — bring your own API key; no data leaves your vault except to the endpoint you configure
• Cross-platform — same bundle runs on desktop (macOS / Windows / Linux) and mobile (iOS / Android)

Installation

From Obsidian Community Plugins (coming soon)

Search for OpenAgent in Settings → Community Plugins → Browse.

Manual install

1. Download main.js, styles.css, and manifest.json from the latest release.
2. Copy them into <vault>/.obsidian/plugins/open-agent/.
3. In Obsidian: Settings → Community Plugins → enable OpenAgent.

Mobile

Sync the plugin folder to your mobile vault's .obsidian/plugins/open-agent/ via Obsidian Sync or any file-sync tool.

Configuration

Open Settings → OpenAgent and fill in:

Privacy & network use

This plugin makes network requests only to the LLM endpoint you configure (e.g. OpenAI, OpenRouter, Ollama, LM Studio). No data is sent to any other server.

With vault tools enabled, the agent may transmit note bodies, paths, frontmatter, and tags to that endpoint — only use endpoints you trust.

Your API key is stored in .obsidian/plugins/open-agent/data.json. If you sync your .obsidian/ folder (e.g. via Obsidian Sync), the key travels with it.

Development

git clone https://github.com/nikitaclicks/obsidian-openagent.git
cd obsidian-openagent
npm install
npm run dev        # esbuild watch → writes main.js

Then copy main.js, styles.css, and manifest.json into a test vault at <vault>/.obsidian/plugins/open-agent/ and enable the plugin.

npm run build      # production bundle
npm run lint       # TypeScript ESLint

Project layout

src/
  main.ts          # plugin entry point
  settings.ts      # settings tab + defaults
  view.ts          # chat panel UI
  loop.ts          # agent run loop
  provider.ts      # OpenAI-compatible HTTP client
  types.ts         # shared types
  tools/
    registry.ts    # tool registration
    vault/         # read / write / edit / search / list / frontmatter / links
  consent/
    manager.ts     # per-action consent state
    modal.ts       # confirmation dialogs
    render-diff.ts # diff previews

Contributing

Pull requests are welcome. For larger changes, open an issue first to discuss what you'd like to change.

1. Fork the repo and create a branch from main.
2. Make your changes with tests where applicable.
3. Run npm run lint and fix any issues.
4. Open a PR — describe what changed and why.

Roadmap

• Additional providers (Anthropic native, Google Gemini)
• MCP server support
• Inline agent commands
• Tool call history / audit log

License

MIT