Hey everyone,

I’ve been using Obsidian for personal knowledge + AI workflows, and I kept running into one core issue:

👉 Obsidian is powerful for humans, but not designed for programmatic automation or AI agents.

⚠️ The problem

When connecting my vault to tools like:

• Claude Code
• Codex
• local LLM agents

I kept hitting:

• messy CLI scripts
• no clean API abstraction
• no structured way to query notes
• no graph-level access
• no real “memory layer” for AI

💡 Insight

Obsidian isn’t just a note app.

It’s actually a local graph-based knowledge system.

But there was no proper developer interface to use it like one.

🧩 So I built this

👉 Obsidian SDK + API Wrapper (npm package)

A TypeScript SDK + universal API layer over the Obsidian CLI.

It turns your vault into a programmable system.

⚙️ What it enables

📌 Local automation
→ create / update / manage notes programmatically

📌 Graph workflows
→ treat notes as connected nodes

📌 AI memory layer
→ persistent context for Claude Code, Codex, local agents

📌 Language-agnostic API
→ usable from any language via API layer

🧠 Why it matters

This turns Obsidian into:

👉 a local-first memory + knowledge graph engine

Instead of rebuilding context every time, your vault becomes:

• memory system
• retrieval layer
• automation backend

🌱 Note

This is my first npm package built solo.

So:

• bugs are possible
• APIs will evolve
• feedback is welcome
• contributions appreciated

🔗 Links

GitHub (SDK): https://github.com/MYSELF-SAYAN/obsidian_cli_sdk
GitHub (API): https://github.com/MYSELF-SAYAN/obsidian_cli_api
npm: https://www.npmjs.com/package/obsidian-sdk
Docs: https://obsidiansdk.vercel.app/

💬 Would love feedback on:

• AI workflows with Obsidian
• graph-based retrieval systems
• local LLM memory setups
• edge cases from power users

If you’ve tried turning Obsidian into an AI memory layer, I’d love to hear how you approached it.