n8n Knowledge Plugin for Claude Code
Live
Stop babysitting web search permissions. This Claude Code plugin replaces 5-10 tool calls with a single 1-second lookup against a 443K-memory knowledge graph. Community solutions, workarounds, official docs, GitHub issues with status — all pulled via graph-based search that connects content by themes and context, not just keywords. Every result is scored by confidence and linked to its source. One-command install, zero config.
Claude Code Pluginn8n DocsDeveloper ToolingZero Config
Case study → Evals & Benchmarking — n8n Workflow Generation
Live
Claims are cheap. I built an eval harness that benchmarks AI-generated n8n workflows head-to-head against the most popular community tool — 128 prompts, multiple models, ~2,800 scored generations — judged by the n8n-mcp validation engine plus a blinded Claude Opus intent-judge. Then I used it as the decision gate for the plugin: every change measured, several reverted because the numbers said no. Leads with honest scope and limitations.
EvalsLLM-as-JudgeBenchmarkingProduction AI
Case study → n8n Pulse — Community Intelligence Chatbot
Live
Claude-powered chatbot: Ask it what *actually* matters most to the community, then what files in the codebase actually need attention for the top issues, and what the likely fixes are. Or let it guide you through a few examples yourself to quickly see how powerful it is. It's powered by 443k+ memories with 12.7M graph links between them. All from n8n's ecosystem — Docs, GitHub issues, forum questions, feature requests, and the source code with themes running across all of them. Native tool calling with budget control, tag filtering, and TEMPR graph recall. Cron jobs keep it up to date with the latest updates nightly.
Claude Tool CallingHindsight Graph RAGCommunity Analytics443K+ Memories
Case study →
Burrfect AI Enhancement Pipeline
Live Two-provider AI pipeline built before the tools existed. Eval harness regression-tests model upgrades.
EvalsGuardrailsMulti-Provider AI
Case study →
Burrfect PMF Dashboard
Live 13-page analytics dashboard backed by 20+ BigQuery views. Cohort retention, activation research, A/B experiment results.
EvalsA/B TestingBigQuery
Case study →
Merge Duplicates Dashboard
Live Admin tool for deduplicating beans/roasters across Firestore and WordPress. Data quality for the recommendation engine.
Data QualityHuman-in-the-LoopFirebase
Case study →
CREOpsDesk
Beta Django platform for CRE marketing coordinators + n8n browser automation for listing syndication. Thin-loader Puppeteer architecture — fix a selector once, every client instance updates.
n8nMulti-TenantPuppeteerDjangoThin-Loader
Case study →
Portfolio AI Assistant — Kumpel 2.0
Live RAG chatbot backed by a Hindsight graph database. The engine behind n8n Pulse — Claude Code hooks, Skills, and an MCP server keep the knowledge base current as site content evolves.
Graph RAGMCPHooksMulti-Provider LLMSSE Streaming
Case study →
Burrfect Water
Live Water chemistry tool for espresso: constraint satisfaction across 10 SCA parameters for 44 European cities. Deterministic math, no LLM guessing.
GuardrailsConstraint SatisfactionData Research
Case study →
Personal Ops Orchestrator
Live The more AI agents I ran in parallel, the harder it was to track their status and stay ahead of a flood of inputs — including foreign-language email as a new-country parent. So I built an orchestrator: overnight, on a local model for privacy, it triages my inbox, translates and prioritizes what matters, reconciles my task list, and leaves one daily brief. Daily production since early 2026.
OrchestrationLocal LLMAutonomous AgentsMCPGuardrails
Case study →
Project Orchestrator
In Dev Two-tier agent system: strategic planner + autonomous executors in Vagrant-sandboxed VMs. TDD enforced at the plan level. The architecture that made safe autonomous AI development possible.
Claude CodeAutonomous AgentsGuardrailsTDD
Case study →
BTC Backtesting Engine
In Dev Walk-forward backtesting with statistical validation, regime-aware evaluation, and overfitting detection. A safe proving ground for autonomous AI development with abundant data and measurable outcomes.
EvalsAutonomous AgentsStatistical Validation
Case study →
Woordjes — Dutch Learning App
Live Multi-user flashcard app with AI-generated word lists, PDF ingestion from a Dutch tutor, and a human-in-the-loop chat for refining vocabulary.
DjangoClaude APIHuman-in-the-LoopGuardrails
Case study →
Meal Planning Pipeline
Live A 3-minute conversation produces a deployed app: shopping lists organized by store layout, cook-mode walkthroughs with timers, published to a family docs server. Skills composition in action.
Claude CodeSkillsHuman-in-the-Loop
Case study →
CRE Field Mapping Skill
Live A "Skill" in n8n's vocabulary. 15 rules from production failures making AI reliable at complex field mapping tasks.
SkillsGuardrails
Claude Slack
Deprecated Full Claude Code power from your phone via Slack. Emoji-based quick responses, speech-to-text input, push notifications when Claude needs attention.
Claude CodeSlack APIMCP
Case study →