The pattern you probably hit
Most people who reach out about this guide had the same five-step sequence:
- Opened a new AI chat inside n8n (the sparkles / "Ask AI" sidebar).
- n8n asked you to pick a model — OpenAI, Anthropic Claude, Google Gemini, etc.
- You picked one. It prompted for an API key.
- You tried to add the key and got the error "your role can not create a key".
- You followed the trail to Settings → Project Roles and were told to upgrade to Enterprise.
This guide explains why none of that is a Dargo configuration problem, where the boundary actually sits, and how to do what you were trying to do using the right n8n feature.
TL;DR
- The "Ask AI" / chat sidebar inside n8n is an n8n Cloud-tier feature. It surfaces in the UI on every edition but is not fully usable on the self-hosted Community Edition our AppStore ships.
- The error "your role can not create a key" is misleading. It sounds like a permissions problem but the underlying reason is feature-tier gating. The Owner role on your install can absolutely create normal credentials — just not this specific Cloud-tied credential type.
- The AI Agent workflow node is a different feature in the same product. It works fine on Community Edition, it's the actual way you should use AI in n8n on a self-hosted box, and step-by-step instructions are below.
- There is no Dargo-side switch that unlocks the AI Assistant. Community users have tried the obvious env vars (
N8N_AI_ENABLED=true,N8N_GENAI_ENABLED=true,N8N_BLOCK_BUILDER_ENABLED=true) and confirmed on n8n's own forum that they don't lift the gate — the check happens server-side on n8n's Cloud service, not in your container.
Two features, very similar names
n8n ships with two "AI" things that share confusingly similar names. The difference matters here.
AI Assistant — the chat sidebar
This is the in-product helper. You open it from a sparkles or chat icon in the n8n UI and describe what you want — "build me a workflow that sends a Slack message when a Gmail arrives" — and it walks you through building the workflow.
Even though you're self-hosting n8n itself, the AI Assistant expects to talk to an LLM that n8n hosts on their side as part of their paid plans. The model picker, the API-key prompt, and the role-gated error are all part of that Cloud flow. On Community Edition the path to that managed service is closed.
That's why "your role can not create a key" is misleading. The real meaning is closer to "this credential type isn't enabled on your plan." n8n surfaces it as a role error because the access check lives in the same code path that gates role-restricted credentials, but functionally, even an Owner can't create the specific credential this feature wants on Community Edition.
The same applies to Settings → Project Roles: the "Projects" feature (multiple isolated projects each with their own role assignments) is Enterprise-only per n8n's own documentation. On Community Edition you have one Personal project and that's it.
AI Agent node — the actual usable feature
This is the one you want.
It's a node you drop into a workflow, not a sidebar. You configure it with an LLM provider of your choice, paste in your own API key from that provider, and the node calls the LLM whenever the workflow runs.
The AI Agent node is part of the open Community Edition. The credential creation works normally because the credential goes directly from your Dargo device to the LLM provider, not through any n8n Cloud service. There's no plan-tier gate.
How to use AI in n8n on your Dargo device, step by step
OpenAI is the example below; the same pattern works for Anthropic, Google AI Studio, Groq, Mistral, local Ollama, and others.
Step 1. Get an API key from the LLM provider
You'll need an account directly with whichever LLM provider you want to use:
- OpenAI:
platform.openai.com→ API Keys → create a new key. Add billing; there's no free tier on the API even for GPT-4o-mini. - Anthropic (Claude):
console.anthropic.com→ API Keys → create. Add credit. - Google AI Studio (Gemini):
aistudio.google.com→ get an API key. Free tier exists, rate-limited. - Groq:
console.groq.com→ free tier, very fast inference of Llama-family models.
Copy the API key to your clipboard or a password manager. You'll paste it into n8n in step 4.
Step 2. Create a workflow
In n8n's left nav, click Workflows (not the AI chat sidebar). Click Create Workflow.
Step 3. Add a Chat Trigger
Add a Chat Trigger node. This gives the workflow a webhook URL you can open in a browser to talk to it. After saving the workflow, the node will display the public URL — that's the chat surface you and your users will hit.
Step 4. Add an AI Agent node
Connect an AI Agent node to the Chat Trigger. The AI Agent node has three sub-slots:
- Chat Model — the LLM provider
- Memory — optional, lets the agent remember earlier turns
- Tools — optional, lets the agent call other workflows or HTTP APIs
Click the Chat Model sub-slot and pick whichever provider you set up in step 1 (OpenAI Chat Model, Anthropic Chat Model, Google Gemini Chat Model, etc).
When the chat model node opens, click the Credentials dropdown → Create new credential. Paste the API key from step 1, give the credential a name, save. This is where the "your role can not create a key" error does not appear, because this is a regular n8n credential, not the Cloud-only AI Assistant credential.
Step 5. Save and activate
Click Save in the top-right of the workflow, then toggle the workflow Active. The Chat Trigger node now exposes a public chat URL (visible in the node's settings panel). Open it in a new tab and start chatting — your messages flow through the AI Agent node, the node calls your LLM with your API key, and replies stream back to the chat UI.
That's the productive n8n self-hosted AI flow. The AI Assistant sidebar is a Cloud-only convenience layer; the actual AI capability lives in workflows.
Why we don't unlock the Assistant from the Dargo side
You may reasonably wonder whether Dargo could enable the Cloud-tier AI Assistant by setting an environment variable in the Docker container we ship. We can't — and we've confirmed why.
The two environment variables that look like they should help are N8N_GENAI_ENABLED and N8N_BLOCK_BUILDER_ENABLED. Multiple self-hosted users have tried setting both to true, verified they loaded correctly inside the container with docker inspect, and reported on the n8n community forum that the feature still doesn't appear. The official documentation page for AI Assistant environment variables links to a Cloud-only path that returns 404 on self-hosted instances.
The short version: the feature gate is enforced server-side on n8n's Cloud service, not by your container's configuration. There is no Dargo-side switch.
So is our installation broken?
No. Specifically:
- The Owner account is set up correctly and can create any normal credential.
- Workflows, the AI Agent node, the Chat Trigger node, integrations with the 400+ external services n8n supports, code nodes, expression evaluation, schedule triggers, webhooks — all of this works.
- The only thing that doesn't work is the Cloud-tier AI Assistant, and that's by n8n's design, not because we configured anything wrong on the container.
If you really want the in-product Assistant chat, the practical options are:
- Skip it. Build workflows in the regular editor. The AI Agent node gives you the actual AI capability, which is what most self-hosters are after.
- Subscribe to n8n Cloud. Separate subscription with n8n directly; unrelated to Dargo.
A note on n8n's broader Community vs Enterprise gating
The AI Assistant isn't the only feature n8n keeps out of the Community Edition. Per the official documentation, the Community Edition omits several capabilities reserved for Cloud and Enterprise plans, including: Projects, custom variables, environments, external secret backends, external binary-data storage, log streaming, multi-main mode, single sign-on (SAML / LDAP / OIDC), and a few others.
If you're picking automation tools for a personal Dargo device and want one that doesn't push paid-tier prompts on self-hosters, Automatisch (also in our AppStore catalog) is a more no-code-friendly alternative. n8n is the more powerful tool for developers and is worth the workflow learning curve; Automatisch is friendlier if you just want Zapier-style trigger-action automations without the upsells.
Summary
- "Your role can not create a key" + the AI chat sidebar + Project Roles asking for Enterprise: all three are facets of the same thing — the AI Assistant is a paid Cloud / Enterprise feature not enabled on Community Edition self-hosted installs.
- The error message is misleading. The Owner role on your install can create normal credentials.
- This is not a Dargo configuration issue, and there is no environment variable we can set to fix it from our end.
- The productive way to use AI in n8n on your Dargo device is the AI Agent workflow node, which works fine on Community Edition with your own LLM provider API key.
References
- n8n Community Edition features — the official list of what's Community vs Enterprise.
- n8n account types and roles — confirms Owner can create credentials and that the Admin/Member account-type split is itself a paid-plan feature.
- Community thread: AI Assistant features missing despite N8N_GENAI_ENABLED=true (Dec 2025) — multiple users confirming env vars don't lift the gate.
- Community thread: Missing Credentials and AI menu items as the Owner user — same pattern, no fix.