Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/SamBleed/opencode-obsidian/llms.txt

Use this file to discover all available pages before exploring further.

Bunker OS is a local-first knowledge operating system built on Obsidian and OpenCode. It turns AI sessions, research, audits, evidence, and decisions into persistent operational assets — with async automation via n8n, fast text retrieval via BM25, and a full orchestration pipeline purpose-built for OpenCode. Version 1.3.1 ships with 430 passing tests, 13 bundled skills, and a complete n8n automation layer including a Dead Letter Queue and multi-channel alerting.

What Bunker OS Is

Bunker OS is not a personal knowledge manager. It is an operations system — a platform for agents and humans to work together on research, security audits, infrastructure decisions, and project intelligence, with everything persisted as plain Markdown in an Obsidian vault on your local disk. The core insight comes from Andrej Karpathy’s LLM Wiki pattern: rather than feeding every AI session the same raw documents, you maintain a persistent, cross-referenced wiki that grows richer with each session. Knowledge compounds like interest. Every autoresearch, every ingest, every think session leaves behind structured artifacts — entities, concepts, source summaries, ADRs — that every future session can draw from without re-reading the originals.

The Operations-First Difference

Bunker OS forked from claude-obsidian but evolved in a fundamentally different direction. While claude-obsidian excels at PKM methodologies (LYT, PARA, Zettelkasten) and multi-writer safety, Bunker OS is built around the operational concerns that matter most to developers and security engineers: automation pipelines, evidence preservation, error handling, and measurable test coverage. Key differentiators from any other AI-Obsidian integration:
  • n8n async automation — a full Docker-based workflow engine with webhooks, AI triage, GitHub issue creation, and Slack/Telegram/Discord notifications. No other comparable tool has this.
  • Dead Letter Queue — a global n8n error trigger that catches failures from every workflow in the instance, classifies severity, and persists the last 200 errors.
  • Evidence vaultreport.zip and security-audit-report.json are indexed with SHA256 checksums so audit artifacts are verifiable.
  • 430 tests across 5 suites — the codebase is CI-tested on every push and PR to main via GitHub Actions.
  • BM25 text retrieval — pure Python stdlib, zero dependencies, no embeddings, no external API calls. The agent is the only intelligence in the retrieval loop.

Architecture Layers

Bunker OS organizes itself into ten conceptual layers that map directly to folders, scripts, and skills in the repository:
LayerWhat it covers
Capturewiki/inbox, .raw/, ingest server, session handovers
Normalize_templates/, skills/wiki-ingest
Classifywiki/sources, wiki/entities, wiki/concepts, wiki/projects, wiki/comparisons
Evidencereport.zip, security-audit-report.json, evidence index with SHA256
GovernanceBUNKER_RULES.md, ADR templates, knowledge supply chain
Automationbin/ for sync ops, n8n for async flows, alerts, and DLQ
RetrievalBM25 index (pure Python, zero deps) + OpenCode synthesis
Testing5 suites: n8n workflows, wiki integrity, scripts, YAML, retrieve
Agent Runtime3 agents, commands, 13 skills, 7 hooks across 4 events
VisualizationDashboard, canvases, Obsidian graph view

Why Bunker OS

Bunker OS sits at the intersection of knowledge management, automation, and operational tooling — a space no other tool covers completely:
CapabilityBunker OSclaude-obsidianSmart Connections
Self-organizing wikiCreates entities, concepts
n8n automationAsync + webhooks + DLQ
Autoresearch3-round web with gaps
Text retrievalBM25 (stdlib, zero deps)BM25 + API
Thinking framework10 principles
Tests + CI430 tests, GitHub Actions
Dead Letter QueueGlobal error handler
Evidence vaultSHA256 checksums
Multi-channel alertsSlack / Telegram / Discord
Health monitoringActive n8n workflow
Multi-methodology❌ (generic only)LYT / PARA / Zettelkasten
Multi-writer safe❌ (single-writer)Per-file locks
Open sourceMITMITFreemium
Bunker OS does not compete with claude-obsidian — it complements it. Where claude-obsidian is a researcher, Bunker OS is an operator. Both can coexist pointing at the same vault.

Who It’s For

Bunker OS is designed for three primary audiences:
  • Developers using OpenCode who want AI sessions to accumulate knowledge rather than restart from scratch each time
  • Security researchers and DevOps engineers who need to preserve audit evidence, track vulnerability decisions, and run automated remediation pipelines
  • Power users who want async workflows and multi-channel alerting wired directly to their knowledge base

Comparison with claude-obsidian

DimensionBunker OSclaude-obsidian
Versionv1.3.1v1.9.2
PurposeOperations + automation + securityPKM / second brain
HarnessOpenCodeClaude Code (plugin)
LanguageEnglishEnglish
Skills13 operational15 research
Async automationn8n on DockerNone
Text retrievalBM25 (delegates to agent)BM25 + API + rerank
Tests430 (5 suites)~1,240 (9 suites)
Evidence checksums
Dead Letter Queue
Multi-channel alertsSlack / Telegram / Discord
Multi-methodologyLYT / PARA / Zettelkasten
Multi-writer safe
Plugin marketplaceClaude Code plugin
Stars8.5k

Explore Bunker OS

Quickstart

Clone the vault, open it in Obsidian and OpenCode, verify hooks, and run your first autonomous research session in minutes.

Requirements

Hardware, software, and optional dependencies — OpenCode, Obsidian v1.6+, Python 3.10+, Docker, and more.

Skills Overview

Explore all 13 bundled OpenCode skills: autoresearch, think, wiki-ingest, wiki-retrieve, canvas, and more.

n8n Automation

Set up the n8n async pipeline with AOC v4 Enterprise, Dead Letter Queue, Health Check, and multi-channel alerting.

Build docs developers (and LLMs) love