Index / n8nClaw · updated Oct 24, 2026
n8nClaw
shabbirun/n8nclaw · warning · rank 42 of 50 by stars
A no-code reimagining of OpenClaw built entirely as a single n8n visual workflow, wiring Telegram, WhatsApp, and Gmail into a Claude-powered agent with Postgres memory and Supabase RAG. It trades OpenClaw's local runtime for drag-and-drop composability on top of n8n's automation ecosystem.
Facts
Repository
Runtime
Posture
Community
Security breakdown
Composite 45 / 100 · how these are scored
higher is safer
higher is safer
higher is safer
higher is safer
higher is riskier
Model access
Read from the repository, not written by a model · 1 files examined
No provider wiring was found in this repository. That is not the same as the project supporting none — it usually means the model call lives somewhere the detector does not read, such as an exported workflow or a binary dependency.
Decision
Why choose n8nClaw over OpenClaw?
Why choose this
- Visual, editable workflow instead of a code runtime
- Leverages n8n's large integration library out of the box
- Built-in multi-channel triggers (Telegram, WhatsApp, Gmail) with minimal config
Tradeoffs
- Depends on n8n plus multiple cloud APIs rather than running locally
- No sandboxing or execution isolation guarantees
- Very young project: two commits, no releases, no license
Best fit
- n8n users who want an OpenClaw-style assistant without writing code
- Teams already running n8n, Postgres, and Supabase infrastructure
- Rapid prototyping of multi-channel AI assistants
Avoid if
- You need a true local-first agent with sandboxed execution
- You want a mature, versioned product with releases and tests
- You avoid depending on multiple external APIs (OpenRouter, Evolution, OpenAI)
Evidence is limited to the README, two initial commits, and generic OpenClaw-ecosystem web results with zero direct mentions of this project. Assessments of security, stability, and community sentiment are therefore low-confidence inferences.
AI layer reviewed Oct 5, 2026 · how this is written
Star activity
230 stars today
Overview
n8nClaw is a lightweight, self-hosted AI assistant implemented entirely as an n8n visual workflow, positioned as a lite alternative to OpenClaw. Instead of a compiled runtime, the whole agent — triggers, routing, memory, and tools — lives in a single importable n8n workflow, making it approachable for no-code and automation-oriented users.
Architecturally, it fans in from multiple channel triggers (Telegram, WhatsApp via Evolution API, Gmail polling, and an hourly autonomous heartbeat), normalizes messages, and routes them through a central AI agent node powered by Claude Sonnet 4.5 via OpenRouter with a 15-message Postgres chat memory. A configurable 'soul' field defines personality, and a living user-profile document personalizes responses. Replies are switched back out over the originating channel.
Where it differs most from OpenClaw is its tool and memory layer: tiered worker sub-agents, task/subtask databases, email and document managers, and a scheduled memory pipeline that aggregates chat history, summarizes it with Haiku 4.5, embeds it with OpenAI, and stores it in a Supabase vector store for RAG. The trade-off is heavy reliance on external services and n8n itself, plus the project is brand new — two commits, no releases, and no declared license — so production readiness is unproven.