Obsidian Agent Client: Bring Claude Code, Codex, and Gemini CLI Into Your Vault

Obsidian Agent Client in One Sentence

Obsidian Agent Client is an open-source Obsidian plugin that lets you run AI coding agents directly inside your vault sidebar, with note-aware context and tool execution.

If you live in Obsidian and jump to terminal/IDE chats all day, this plugin closes that loop.

What It Actually Does

Agent Client embeds a chat panel in Obsidian and connects it to agents that support the Agent Client Protocol (ACP).

You can:

Chat with Claude Code, Codex, Gemini CLI, or other ACP-compatible agents
Mention notes with @notename to pass note context to the agent
Use slash commands and keyboard workflows
Let agents run terminal commands and return results inline
Control file access/editing with permission flows
Export chats into your vault for long-term memory

Why This Is Different From Generic AI Chat Plugins

Most Obsidian AI plugins are provider-specific or API-only.

Agent Client is more like a protocol bridge:

It is agent-agnostic (as long as the agent supports ACP)
It supports multi-agent switching from the same interface
It focuses on developer workflows (terminal, edits, sessions, model/mode controls)
It keeps the interaction anchored in your vault context

Supported Agents (Current)

From the project docs, the core supported setups are:

Claude Code (Anthropic, via Zed ACP adapter package)
Codex (OpenAI, via Zed ACP adapter package)
Gemini CLI (Google, ACP mode)
Custom ACP-compatible agents (for example OpenCode/Qwen Code)

Installation (Current Path)

As of February 18, 2026, the docs state the plugin is still under review for the Obsidian Community Plugins directory, so recommended install is via BRAT.

Option 1: BRAT (Recommended)

Install BRAT from Community Plugins.
Add beta plugin repo URL: https://github.com/RAIT-09/obsidian-agent-client
Enable Agent Client.

Option 2: Manual

Download main.js, manifest.json, styles.css from releases.
Copy into: VaultFolder/.obsidian/plugins/agent-client/
Enable plugin in Obsidian.

Typical Setup Flow

Install Node.js (only needed for npm-based agents).
Install your agent package/binary.
Find absolute path (which on macOS/Linux, where.exe on Windows).
Put paths + auth info in Settings -> Agent Client.
Open chat panel and send a test message.

On Windows, the docs recommend WSL mode for smoother compatibility.

Key Advantages

One workspace: Notes + AI coding agent in the same UI.
Context precision: @ mentions reduce accidental full-vault exposure.
Multi-agent flexibility: Swap providers/models without changing apps.
Practical for iterative writing/coding: Edit, ask, run, export, repeat.
Open source + Apache-2.0: Easy to audit and adapt.

Tradeoffs and Risks

Early-stage friction: Setup is more technical than one-click SaaS chat.
Path/auth issues are common: Especially on Windows, Flatpak Linux, and Homebrew path differences.
Permission design matters: Agent file edits and shell execution need strict boundaries.
Not officially affiliated with model providers: It integrates their tooling, but remains community-developed.

Who Should Use It

Use Agent Client if you:

already manage serious notes/workflows in Obsidian,
want coding agents in-context without constantly app-switching,
and are comfortable with CLI setup and debugging.

If you want zero configuration and no local path management, this may feel too hands-on right now.

Final Take

Obsidian Agent Client is one of the more promising “AI + knowledge base” bridges because it treats Obsidian as a working surface, not just a note database.

It is best viewed as an advanced plugin for technical users today, but the architecture (ACP + multi-agent switching + vault-aware context) gives it a strong long-term direction.