Skip to main content
Projects

Sidecar AI

Native AI chat sidebar for Obsidian supporting Google Antigravity, GitHub Copilot, Pi Coding Agent, and local AI CLI tools with zero API keys.

Sidecar AI

Overview

Sidecar AI brings modern AI coding agents and LLM reasoning directly into Obsidian as a native sidebar chat panel. Instead of requiring third-party API keys, billing setups, or monthly token allowances, Sidecar AI hooks directly into your already-authenticated command-line tools, including Google Antigravity (agy), GitHub Copilot (copilot), Pi Coding Agent (pi), or custom local CLI agents.

It offers an unbundled, privacy-first interface designed to work alongside your daily thinking, research, and technical note-taking without window switching.

Why CLI-Powered Instead of Cloud API Keys?

Most Obsidian AI plugins require individual OpenAI, Anthropic, or OpenRouter API keys. This introduces unnecessary friction: managing credit balances, unpredictable per-token billing, rate limiting, and storing secret API tokens in local vault configuration files.

Sidecar AI flips the integration model: if you already have authenticated developer subscriptions or CLI tools running on your workstation, Sidecar bridges Obsidian directly to those local processes.

  • Zero Per-Token Fees: Leverage the subscriptions (like GitHub Copilot or Google Antigravity) and local LLM runners you already maintain.
  • No API Keys in Vault: No secrets stored in Obsidian settings or synced across devices.
  • Strict Local Privacy: Prompts and vault contents run directly through your local CLI binaries. No intermediate proxy servers intercept, log, or train on your notes.
  • Agentic Capabilities: Take advantage of local tools that support multi-step reasoning, file lookups, and specialized coding workflows.

Related Blog Posts & Essays

I have written in-depth about the underlying philosophy, practical workflows, and local hardware setups behind Sidecar AI:

  • Improve Yourself with AI
    How I use Obsidian voice dictation alongside AI self-reflection to separate fleeting emotions from underlying facts, identify recurring patterns, and stay intentional as a parent.

  • Further thoughts on how to use your knowledge with AI
    A breakdown of how I structure personal notes and technical knowledge in Obsidian, and the two primary patterns I use to query private data with AI without cloud API keys.

  • Second use of local LLMs
    Why having sufficient RAM on local workstations makes running local LLM agents at home practical, private, and cost-effective without cloud subscriptions.