obsidian-second-brain
Persistent agent memory in an Obsidian vault: saving decisions, search, research and background vault upkeep
Install
/plugin marketplace add eugeniughelbur/obsidian-second-brain
/plugin install obsidian-second-brain@obsidian-second-brainAdd "env": {"OBSIDIAN_VAULT_PATH": "/path/to/your/vault"} to ~/.claude/settings.json and restart Claude Code.
This is third-party code. Review the repository files before installing.
What it does
The plugin turns an Obsidian vault into shared memory for Claude Code and other agents. The save command splits a conversation into decisions, people, tasks and ideas and files them into the right notes, while search and context loading spare you re-explaining projects each session. New sources, from links and PDFs to voice memos and screenshots, rewrite existing pages, and a separate command reconciles contradictions. A research toolkit uses Perplexity, Grok, YouTube and free sources, and background agents handle a morning brief, nightly cleanup and weekly review.
Who it is for. For developers, founders and researchers who keep notes in Obsidian and want the agent to remember context across sessions.
Good fit when
- You are tired of re-explaining projects and decisions to the agent
- You need meeting and conversation outcomes saved as structured notes
- You want topic research collected straight into your knowledge base
Not a fit when
- You do not want the agent rewriting existing notes on its own
- The vault holds data that must not go to external APIs
Example request
Save the outcomes of this meeting to the vault: decisions, people and tasksLimitations
Needs git and uv, plus jq for the script install. Core commands work without keys, but X, Perplexity, NotebookLM and extended YouTube need xAI, Perplexity, Gemini and YouTube keys, some paid. Audio transcription installs Whisper with PyTorch. There are many commands, namespaced in plugin mode.
How to disable. Remove the plugin via /plugin and delete OBSIDIAN_VAULT_PATH from ~/.claude/settings.json. For the script install, delete ~/.claude/skills/obsidian-second-brain and the registered hook.
Security check
- Rewrites vault notes and runs background agents
- Sends requests to external APIs when research commands are used
README in short
The README describes context loss between sessions and contrasts the approach with Karpathy's LLM Wiki: sources rewrite pages instead of only appending. It walks through command scenarios and a before and after table. Claude Code installs via the plugin marketplace or a script, other agents via per-platform builds. It ends with optional research API keys and their costs.
FAQ
Do I need API keys?
Not for core vault commands. Keys are only needed for some research commands, and /research falls back to free sources without one.
Does it work with other agents?
Yes, build.sh produces builds for Codex, Gemini CLI, OpenCode and others.
Related
A CLI for every Google Workspace API with JSON output and agent skills: Drive, Gmail, Calendar, Sheets and more
Local search over Markdown notes, docs and meeting transcripts: keywords, semantic search and reranking, with an MCP server
A task manager for AI-driven development: breaks a PRD into dependent tasks and guides the agent through them via MCP or CLI
Anthropic's official plugins for knowledge work roles: sales, support, product, marketing, legal, finance, data and enterprise search