Bedrock

Turn any Obsidian vault into a structured Second Brain with AI agents

Bedrock is a Claude Code plugin that automates Obsidian vault management through AI-powered skills. It organizes knowledge into 7 entity types following adapted Zettelkasten principles — entity detection, bidirectional linking, ingestion from external sources, deduplication, and sync.

No build system. No runtime. Just markdown files, AI agents, and your Obsidian vault.

Features

• 6 AI-powered skills — query, teach, preserve, compress, sync, and setup
• 7 entity types — actors, people, teams, topics, discussions, projects, and fleeting notes
• External source ingestion — Confluence, Google Docs, GitHub, CSV
• Bidirectional wikilinks — automatic cross-referencing with Obsidian graph view
• Hierarchical tags — multi-dimensional filtering (type/, status/, domain/, scope/)
• Zettelkasten structure — permanent, bridge, index, and fleeting note roles
• Trunk-based git workflow — structured commit conventions built in

Installation

/plugin marketplace add iurykrieger/claude-bedrock
/plugin install bedrock@claude-bedrock

For local development:

claude --plugin-dir ./claude-bedrock

Quick Start

After installing, run the setup wizard:

/bedrock:setup

This will guide you through:

1. Language selection — choose the vault content language (default: English)
2. Dependency check — verify graphify is installed (required)
3. Vault objective — pick a preset (engineering team, product management, company wiki, personal second brain, open source project, or custom)
4. Scaffold — create directories, templates, config, and connected example entities

The setup creates all entity directories, copies templates, generates a vault-level CLAUDE.md, and scaffolds example entities with bidirectional wikilinks so you can see the graph in Obsidian immediately.

Skills

Vault Structure

your-vault/
├── actors/          # Systems, services, APIs (permanent notes)
├── people/          # Contributors, team members (permanent notes)
├── teams/           # Squads, organizational units (permanent notes)
├── topics/          # Cross-cutting subjects with lifecycle (bridge notes)
├── discussions/     # Meeting notes, conversations (bridge notes)
├── projects/        # Initiatives with scope and deadline (index notes)
└── fleeting/        # Raw ideas, unstructured captures (fleeting notes)

Each directory contains a _template.md defining the frontmatter schema for that entity type.

How It Works

Bedrock turns your vault into a living knowledge graph by combining 8 skills you invoke from Claude Code. You never write entities by hand — skills detect, create, and link them for you, with Obsidian rendering the result as a graph.

First-time use

1. Open a folder you want to turn into a vault (or an existing Obsidian vault).
2. Run /bedrock:setup — answers a few questions and scaffolds directories, templates, and example entities.
3. Open the folder in Obsidian. You'll already see a connected graph.

Day-to-day loops

• Capture knowledge from a source — paste a Confluence page, Google Doc, GitHub repo, or local file into /bedrock:teach. Bedrock extracts entities and writes them to the vault with bidirectional links.
• Ask the vault questions — use /bedrock:ask for anything like "who owns the billing API?" or "what's the status of project X?". It searches the graph, follows wikilinks, and answers with citations.
• Keep sources fresh — run /bedrock:sync to re-pull external sources, or /bedrock:sync --github / --people to surface recent activity and contributors.
• Clean up drift — run /bedrock:compress to fix broken backlinks, merge duplicates, and consolidate fragmented concepts. Run /bedrock:healthcheck for a read-only report.
• Manage multiple vaults — register several vaults with /bedrock:vaults; target a specific one with --vault <name>.

What you get in Obsidian

Every entity has YAML frontmatter (type, status, domain, sources), hierarchical tags (type/actor, status/active, domain/payments), and bidirectional wikilinks. The graph view becomes a navigable map of people, systems, teams, topics, and projects — updated automatically as you teach Bedrock new content.

Dependencies

Confluence and Google Docs ingestion are built into the plugin as internal skills (/bedrock:confluence-to-markdown, /bedrock:gdoc-to-markdown) invoked by /bedrock:teach and /bedrock:sync — no external installation required.

Configuration

Configuration is stored in .bedrock/config.json inside your vault. Run /bedrock:setup again at any time to reconfigure.

Contributing

Contributions are welcome! Here's how to get started:

1. Fork the repository
2. Clone your fork and install the plugin locally:

   claude --plugin-dir ./claude-bedrock
   

3. Create a branch for your feature or fix
4. Make your changes — skills live in skills/, entity definitions in entities/, templates in templates/
5. Test by running the plugin against a test vault
6. Open a PR against main

Project Structure

claude-bedrock/
├── .claude-plugin/    # Plugin manifest (plugin.json)
├── skills/            # Skill definitions (SKILL.md per skill)
│   ├── setup/
│   ├── query/
│   ├── teach/
│   ├── preserve/
│   ├── compress/
│   └── sync/
├── entities/          # Entity type definitions
├── templates/         # Frontmatter schema templates
├── docs/              # Documentation assets
├── CLAUDE.md          # AI agent instructions
└── README.md

License

MIT — Iury Krieger