Le Répertoire des AgentsFR · EN · ES · JA · HI

“I want to connect my apps together without writing code”

Tinkerers juggling ten tools: when a client fills out a form, they need to be added to the CRM, flagged on Slack, and sent an invoice. Doing that by hand, ten times a day, is madness.

Recommended agent: n8n

💰 €0 to €50/month (free self-hosted, ~€20/month in the cloud to start). Level: Intermediate 🌍 Germany Free self-hosted, no time limit

n8n (n8n.io), the visual automation platform with built-in AI nodes. You assemble your automations like Lego: a trigger, a few steps, and your apps talk to each other on their own.

How it works

Trigger: An event occurs in an application — reasons over the data, chains the actions, and logs the result, and you get a multi-tool process run with no manual work.

Step-by-step setup

Create an n8n Cloud account (or install the free version on your server).
Choose a trigger: “when a new message arrives in this form.”
Add steps by connecting them visually (CRM, email, Slack...).
Test with fake data, then activate.

⚠️ Guardrails: Always test with fake data before activating: a badly configured loop can send 200 emails by mistake. Set caps on paid API calls to avoid nasty surprises on the bill.

🔧 Technical details

Installation: hybrid — n8n Cloud (hosted by n8n, no installation) or self-hosted on your own infrastructure.
Prerequisites: Self-hosted: Docker / Docker Compose or npm with Node.js 20.19 to 24.x; SQLite database by default or PostgreSQL (versions 16 to 18 supported); queue mode additionally requires Redis and PostgreSQL. Cloud: just an n8n Cloud account.
Hardware: No general quantified minimum for classic self-hosting; the documentation gives an illustrative sizing (OEM deployment) of 320 MB to 2 GB of memory and notes about 100 MB at rest for a Cloud instance. Scaling goes through queue mode: one main instance, Node.js workers, and Redis.
Self-hosting: yes — documented modes: Docker, Docker Compose, npm (deprecated from n8n 3.0) and Kubernetes (official GKE guide and official n8n-io/n8n-hosting repository).
License: source_available — Sustainable Use License (“fair-code” license created by n8n in 2022) for the repository's main code; “.ee.” files fall under the n8n Enterprise License. n8n explicitly states it is not open source in the OSI sense, since open-source licenses cannot include usage limitations.
API & integrations: API: Yes · SDK: Not documented · CLI: Yes · Webhooks: Yes · MCP: Yes
Models: Documented Chat Model nodes for many providers: OpenAI, Anthropic, Google Gemini and Google Vertex, AWS Bedrock, Microsoft Foundry (Azure OpenAI), Cohere, DeepSeek, Groq, Mistral Cloud, OpenRouter, xAI Grok, among others — each with its own credentials (BYOK; on n8n Cloud, a “Gateway credits” alternative with no provider account or key). Local models supported via Ollama (Ollama Chat Model and Embeddings Ollama nodes).
Data: Self-hosted, workflow data stays on the user's infrastructure (n8n states it is neither a data controller nor a processor); pseudonymous telemetry, on by default (usage, errors, workflow shapes — never workflow data or credential contents), can be disabled (N8N_DIAGNOSTICS_ENABLED=false). On n8n Cloud, data is hosted on Microsoft Azure in Frankfurt (Germany, EU) and telemetry cannot be disabled.
Documented limits: Cloud (documentation): memory per plan — 640 MiB (Starter and Pro-1, 10,000 executions), 1,280 MiB (Pro-2, 50,000 executions), 4,096 MiB (Enterprise); storage up to 100 GB per instance; pruning of stored executions (Starter: 2,500 and 7 days; Pro: 25,000 and 30 days; Enterprise: 50,000 and unlimited duration). Self-hosted, n8n does not limit the amount of data processed per node: the practical limit is available memory, with documented “out of memory” errors.
Operations: Standard logging included; log streaming, SSO (SAML, LDAP, OIDC), RBAC roles and projects, Git version control, and environments are paid features (Business/Enterprise self-hosted, higher Cloud plans), absent from the free Community edition. Documented monitoring: Prometheus metrics and OpenTelemetry traces.

Official sources: docs.n8n.io · docs.n8n.io · docs.n8n.io · docs.n8n.io · docs.n8n.io · docs.n8n.io — verified on November 1, 2026.

Find your agent with the quiz →

Frequently asked questions

How much does n8n cost?

€0 to €50/month (free when self-hosted, ~€20/month in the cloud to start). Free when self-hosted, with no time limit.

Is n8n free?

Yes. Free when self-hosted, with no time limit.

Is n8n hard to get started with?

Level: Moderate. Setup takes 4 steps; the first one: Create an n8n Cloud account (or install the free version on your own server).

What limits and guardrails should you know about with n8n?

Always test with fake data before activating: a badly configured loop can send 200 emails by mistake. Put caps on calls to paid APIs to avoid any nasty surprises on the bill.