Agent

An AI agent for your vault. Chat with it in a sidebar or a dedicated tab, and it reads, searches, creates and edits your notes through 27 built-in tools. It speaks the OpenAI-compatible API, so you can use any provider and model you like — OpenAI, DeepSeek, Moonshot, Ollama, LM Studio and more. Works on both desktop and mobile, and it is free forever.

Features

• 27 function-calling tools over the vault API: read/list/search notes, create/append/overwrite/delete/rename files, metadata and backlinks, workspace commands, live editor selection/cursor edits, and dedicated YAML frontmatter operations.
• OpenAI-compatible client: any /chat/completions style endpoint works. Requests go through Obsidian's requestUrl, so there are no CORS issues.
• Agent loop: the model calls tools, results are fed back, and the loop continues until the model answers in plain text (iteration cap adjustable, optional unlimited mode).
• Long-term memory & skills: an AGENT/ folder is created in your vault root with memory.md and skills/. Memory is injected into every conversation; skill files can be loaded by the agent on demand.
• Two chat modes: right sidebar (ribbon icon) or a standalone tab in the main area.

Installation

From the community plugin directory (recommended)

1. Open Settings → Community plugins in Obsidian and turn off restricted mode.
2. Click Browse, search for Agent, then click Install and Enable.

Manual installation

1. Download main.js, manifest.json and styles.css from the latest release.
2. Create a folder <your-vault>/.obsidian/plugins/agent/ and copy the three files into it.
3. Restart Obsidian, then enable Agent under Settings → Community plugins.

Usage

1. Open the plugin settings and fill in Base URL, API key and Model (e.g. DeepSeek: https://api.deepseek.com/v1 + deepseek-chat).
2. Open the chat panel:
   • Click the robot ribbon icon (opens according to the Chat open mode setting), or
   • Run the command Open chat in sidebar / Open chat in a new tab from the command palette.
3. Type your request and press Enter. Type @ to fuzzy-match and reference vault notes — referenced files' contents are attached automatically. Tool calls and their results are shown inline in the chat.
4. Try things like:
   • "List all notes in the vault root and summarize my note structure"
   • "Find every note mentioning 'deadline' and compile a TODO table into Tasks.md"
   • "Show me the backlinks of the currently open note"

Settings

• Chat open mode: where the ribbon icon opens the chat (sidebar or tab).
• Require confirmation: ask before any vault-modifying tool runs (with diff preview for overwrite/replace).
• Truncate long files + Max lines per read: manual switch and line threshold for how much file content is uploaded per tool call.
• Max tool iterations: safety cap on tool-call rounds per message (1–30).
• Unlimited mode (超限模式): removes the cap; the agent keeps calling tools until the model stops on its own.

Safety

overwrite_note and delete_file modify your vault directly. The agent is instructed to prefer small, safe edits, but please evaluate model behavior yourself and back up important notes.

中文说明

将 Obsidian API 封装为 27 个 LLM function-calling 工具，内置纯 JS（零依赖）的 OpenAI 兼容 agent，以聊天方式驱动它操作你的笔记库，桌面端与移动端体验一致。

使用

1. 设置页填入 Base URL / API Key / Model（如 DeepSeek：https://api.deepseek.com/v1 + deepseek-chat）。
2. 点击 ribbon 机器人图标，或用命令面板执行 “Open chat in sidebar” / “Open chat in a new tab” 打开聊天面板。
3. 首次启用会自动在库根目录创建 AGENT/memory.md 与 AGENT/skills/；记忆全文与技能列表会自动注入 system prompt。

安全提示

overwrite_note / delete_file 会直接修改你的库。agent 受 system prompt 约束（优先小改动），但请自行评估模型行为，重要笔记先备份。