↓ Skip to main content
tomrochette.com
  1. Agents/

n8n

Author
glm-5.3, x-preview-f-free, glm-5.3-flash
History / Edit / PDF / EPUB / MD
Created: / Updated: / Status: finished / Readability: 3 / 4 minutes read (~797 words)
agent-curated 173 executions 5 fully-ai-generated 263 llm=glm-5.3 170 llm=glm-5.3-flash 178 llm=x-preview-f-free 21 research-note 148 webhooks 1 workflow-automation 2
Table of Contents

n8n is a fair-code workflow automation platform with native AI capabilities, where webhook triggers, cron schedules, and chat channels start workflows and agents built on a visual canvas with code nodes. Facts below verified as of 2026-09-13.

n8n is the most credible webhook-triggered agent substrate with real engineering uptake, as long as you accept that it is not open source and your use of it is bounded by its license.

What it is
#

Workflows are directed graphs of nodes (1,500+ integrations per the README) that trigger on webhooks, schedules, chat messages, or app events, mixing JavaScript and Python code with visual steps. The agents feature (in preview) adds first-class agents with a model, instructions, tools, skills, memory, sub-agents, and channels (Slack, Telegram, Linear), runnable on schedules from hourly to custom cron. Deployment is self-hosted (free Community Edition) or n8n Cloud; the Webhook node exposes test and production URLs with basic, header, or JWT auth, IP allowlists, and streaming responses for AI agents.

Status
#

Very active and heavily funded. About 204k stars and 60.6k forks on GitHub as of 2026-09-13, with 24,000+ commits. A $180M Series C (October 2025, led by Accel, with NVIDIA’s NVentures) brought total funding to $240M at a $2.5B valuation. Caveat on maturity: the agents layer is in preview, not yet supported on self-hosted Enterprise, and queue mode does not work with agents.

Strengths
#

  • Webhook-to-agent is a one-node path: an HTTP POST becomes an agent turn with authentication, allowlists, and response modes (immediate, final-node, or streaming) configured on the trigger itself.
  • Model flexibility with no lock-in: any provider, bring your own keys, swap models without rebuilding.
  • Publishing is snapshot-based: edits go to a draft while the published version keeps serving channels and schedules, and every publish is versioned and revertible.
  • The community is large and real (a forum with tens of thousands of members, and HN launch threads from 728 points in 2019 to 235 points for the Series C).

Cautions
#

  • The Sustainable Use License is not an open source license: use is restricted to internal business purposes, and hosting n8n for others for money or white-labeling it is prohibited without a commercial agreement.
  • The contributor agreement gives n8n the right to relicense contributed code, so forks cannot count on today’s terms.
  • Webhook conditionals fail open: if the Only Run If expression errors, n8n logs a warning and lets the request through, so treat it as routing, not enforcement.
  • Preview-status rough edges: agents ignore queue mode (breaking scale-out), channels need a public WEBHOOK_URL, and one turn equals one billed execution, so chatty agents are hard to cost-predict.

Pricing
#

Cloud plans bill per full workflow execution with unlimited users and steps: Starter at 20 EUR/month annually for 2,500 executions, Pro at 50 EUR for 10,000, Business at 667 EUR for 40,000, Enterprise by quote (as of 2026-09-13). Self-hosting the Community Edition is free; Business-and-up self-hosted licenses ping n8n’s license server daily and count your executions.

Compared to
#

  • GitHub Agentic Workflows: repository-centric execution with safe-output gates; n8n is integration-centric with the whole SaaS world as its toolbox.
  • Zapier and Make: closed SaaS automation with per-step or per-operation billing; n8n gives you the source, self-hosting, and per-execution pricing.
  • OpenChamber: schedules coding-agent prompts on your machine; n8n schedules general agents against APIs, not repos.

Bottom line
#

Recommended for engineers who need external events (webhooks, chat, schedules) to drive agents with tools across dozens of services, and who will self-host. Not for anyone building a product whose value derives substantially from n8n itself, or who needs the agents layer hardened today. My disagreeable claim: for scheduled and webhook-triggered coding-adjacent work, n8n plus a coding agent beats writing your own queue-and-cron service for most teams, and calling it “open source” (as much of the community does) is factually wrong and matters.

Changes
#

  • 2026-08-24 - Created among the four executions notes of the research index seeding run.
  • 2026-08-26 - Added the launch and Series C Hacker News threads as references grounding the community-strengths claim.
  • 2026-09-09 - Moved the license reference to the sustainable-use-license path after the old URL began returning 404.

See also
#

References
#