n8n

n8n is what your team builds on once they realise Zapier's $1,500 monthly bill is paying for a runtime they cannot inspect, version, or extend. It is open-source, self-hostable, and built around a node graph that handles AI agents, webhooks, multi-step transformations, parallel branches, and real error handling. We run it ourselves. We build on it for clients. We treat it as production infrastructure, not a toy.

What n8n Does

n8n is a workflow automation platform. You build flows visually as a graph of connected nodes, drop in code where the visual layer runs out of road, and run the whole thing on n8n Cloud or your own infrastructure. It ships with 500+ native integrations and a first-class AI sub-graph for LLM agents, tool use, and retrieval.

• Visual node-based workflow builder with branches, loops, sub-workflows, and parallel execution paths.
• Native AI nodes: LLM chains, agents, tools, vector stores, memory, MCP support, and human-in-the-loop approvals.
• JavaScript and Python code nodes for the moments when no native integration fits. Real code, not a hobbled expression language.
• Webhook triggers, scheduled triggers, polling triggers, and queue-mode execution for high-throughput jobs.
• Self-hostable on your own VPS, Kubernetes cluster, or n8n's managed cloud. Same workflow JSON either way.
• Version control via workflow JSON export plus a Git integration on paid tiers. Workflows are real artefacts you can review.
• Execution history, replay, structured logging, and error workflows. The observability layer Zapier never built.

n8n's AI

n8n's AI layer is where it pulls ahead of every other automation tool we have tested. You can compose multi-agent flows, attach tool nodes that hit any of the 500+ integrations as callable functions, plug in OpenAI, Anthropic, Gemini, or a self-hosted Ollama model, and route through MCP servers. Structured outputs, evaluation harnesses, and human approval gates are first-class nodes. The result is agents that work on real systems with audit trails. Not chat demos that fall over the first time production data hits them.

Automations We Build with n8n

We use n8n as the orchestration backbone for almost every Moonira build. It sits between the CRM, the data warehouse, the email infrastructure, and the inbox, routing events, triggering AI calls, and writing back to the systems your team actually works in. A few of the plays we run repeatedly:

• Inbound lead routing. Form, chat, or webhook fires, n8n enriches via Clay or Apollo, scores against your ICP, dedupes against HubSpot or Salesforce, then assigns the right owner and posts to the Slack channel where deals actually get worked.
• AI inbox triage. Gmail or Outlook messages flow through n8n, get classified by an LLM agent, drafted replies are queued for human approval, and meeting requests are routed to a scheduler. Handles 80-90% of inbound without the AE ever opening the inbox.
• Stripe-to-finance reconciliation. New charges, refunds, and disputes flow into Postgres, get reconciled against QuickBooks, and any variance over a threshold lands in the CFO's Slack with a one-click approve-or-flag.
• Customer onboarding sequences. Closed-Won in the CRM kicks off provisioning in your product, creates the Stripe subscription, sends the onboarding email cascade, schedules the kickoff call, and opens the project in ClickUp or Asana. No handoff dropped.
• Internal AI agents. Voice-of-customer summarisation from Fireflies and Gong transcripts, weekly digest reports written by an LLM and posted to leadership channels, churn-risk scoring that pulls from Stripe, support tickets, and product usage.
• Outbound enablement. Smartlead and Instantly webhooks pipe replies through n8n, classify intent with an LLM, push positive replies to the CRM with full thread context, and silence the noise for the SDR team.

Why Teams Choose n8n

• Execution-based pricing. One workflow run is one execution, no matter how many steps. Teams running 50-step AI agents pay the same per run as a two-step zap. Zapier-style task pricing breaks at scale; n8n does not.
• Self-hostable. Run it on your own infra, behind your own firewall, with your own data. Critical for finance, healthcare, and any team with a real security review process.
• Real observability. Every execution is logged, replayable, and inspectable. When a workflow fails at 3am you can see exactly which node, which input, and rerun from the failure point.
• Code where you need it. The visual layer covers 80% of work, the JS and Python nodes cover the other 20%. You never hit the wall that ends Zapier projects.
• First-class AI primitives. Agents, tools, memory, vector stores, and MCP are native nodes, not bolt-ons. AI workflows feel like production software, not a prompt-engineering experiment.

n8n integrates natively with HubSpot, Salesforce, Slack, Gmail, Postgres, Supabase, Stripe, Shopify, OpenAI, Anthropic, Notion, Airtable, ClickUp, and roughly 500 other systems, plus an HTTP node for anything with an API. Cloud plans start around 20 EUR a month for low-volume teams and scale to mid-three-figures for serious throughput; self-hosting is free, with paid tiers for enterprise features (verify the current public price grid before any commitment). That stack is the build we do. If your team is paying Zapier or Make a few thousand a month and still feels like the automation layer is brittle, that is the conversation to have with Moonira.