n8n Not Working?
AI agent node errors, webhook not receiving, execution timeout (Cloud 5-min limit), self-hosted memory OOM, or LLM credential failures? Check live status and fix it fast.
n8n Cloud — live status
Updated every 5 minutes. Full history at prismix.dev/service/n8n. Self-hosted n8n availability depends on your server.
What's wrong? Diagnose fast
AI agent node errors
Checklist: (1) AI Agent needs at least one Tool connected (use sub-items panel under the node); (2) add LLM credential in Settings > Credentials; (3) exact model names — gpt-4o-mini (not gpt-4-mini), claude-3-5-sonnet-20241022; (4) increase Max Iterations (default 10); (5) add System Message to guide agent behavior.
Webhook not receiving requests
Webhook checklist: (1) activate workflow (green toggle) for production Webhook node; (2) self-hosted: set WEBHOOK_URL=https://your-domain.com in .env; (3) check Executions tab — execution might appear but be erroring; (4) copy the production URL from Webhook node (not test URL); (5) HTTPS required for Stripe, GitHub, etc.
Execution timeout
n8n Cloud: 5 min free, 30 min Pro, unlimited Enterprise. Self-hosted: set EXECUTIONS_TIMEOUT=3600 in environment. For AI agents: split into sub-workflows via Execute Workflow node. Add loop exit conditions to prevent agent infinite loops consuming the entire timeout.
Self-hosted memory OOM
Docker default = 1GB RAM. AI agent + LangChain nodes use more. Fix: docker run --memory 4g. Set N8N_EXECUTIONS_DATA_PRUNE=true + N8N_EXECUTIONS_DATA_MAX_AGE=168 to auto-clean old execution logs. For production: switch from SQLite to PostgreSQL with N8N_DB_TYPE=postgresdb.
LLM credential not working
Go to Settings > Credentials > Add credential > select provider (OpenAI API, Anthropic, etc.). Enter API key. Connect credential to AI node in the node settings sidebar. Common mistake: entering key in wrong field — OpenAI key goes to "OpenAI API" credential type, not "HTTP Header Auth". Credential scoped to workspace, not individual workflow.
Node connection errors / missing data
If nodes show "No data: Execute previous nodes first": workflow has disconnected nodes or depends on webhook input. Test mode: provide sample data via "Edit Fields" node at start. Production: check that upstream trigger node is firing. Common: Webhook trigger fires but HTTP Method mismatch (GET vs POST) causes empty $json.
Key n8n environment variables
Self-hosted .env / docker-compose.yml
# Required for webhooks to work (set your public domain) WEBHOOK_URL=https://n8n.yourdomain.com # Increase execution timeout (seconds, default=0=unlimited on self-hosted) EXECUTIONS_TIMEOUT=3600 EXECUTIONS_TIMEOUT_MAX=86400 # Prune old execution data to save disk/memory N8N_EXECUTIONS_DATA_PRUNE=true N8N_EXECUTIONS_DATA_MAX_AGE=168 # hours (7 days) # Use PostgreSQL instead of SQLite for production N8N_DB_TYPE=postgresdb DB_POSTGRESDB_HOST=localhost DB_POSTGRESDB_DATABASE=n8n DB_POSTGRESDB_USER=n8n DB_POSTGRESDB_PASSWORD=n8n # Tunnel URL for local development (alternative to setting up HTTPS) N8N_TUNNEL=true
Docker run with AI workloads (more memory)
docker run -d \ --name n8n \ --memory 4g \ # increase from default 1g for AI workflows -p 5678:5678 \ -e WEBHOOK_URL=https://n8n.yourdomain.com \ -e EXECUTIONS_TIMEOUT=3600 \ -e N8N_EXECUTIONS_DATA_PRUNE=true \ -v ~/.n8n:/home/node/.n8n \ n8nio/n8n
Correct model names for n8n AI nodes
# OpenAI models (use exact IDs) gpt-4o # correct — NOT "gpt-4-o" gpt-4o-mini # correct — NOT "gpt-4-mini" gpt-4-turbo # correct # Anthropic models (use exact IDs including date suffix) claude-3-5-sonnet-20241022 # correct claude-3-5-haiku-20241022 # correct claude-3-opus-20240229 # correct # Groq models llama-3.3-70b-versatile # correct llama-3.1-8b-instant # correct
Step-by-step fix
- 1
Check live n8n Cloud status
Visit prismix.dev/service/n8n. Also n8n-cloud.statuspage.io. Self-hosted: check your server health directly.
- 2
Fix AI agent node
Connect at least one Tool via the sub-items panel. Add LLM credential in Settings → Credentials. Use exact model IDs (gpt-4o-mini, claude-3-5-sonnet-20241022). Increase Max Iterations from default 10. Add System Message field.
- 3
Fix webhook not receiving
Activate workflow (green toggle, top-right). Self-hosted: set
WEBHOOK_URL=https://your-domain.com. Copy production URL from Webhook node (not test URL). Check Executions tab — if execution exists, webhook arrived but workflow errored. - 4
Fix execution timeout
Cloud: upgrade plan for longer timeouts (Pro = 30 min). Self-hosted:
EXECUTIONS_TIMEOUT=3600. Split long workflows into sub-workflows with Execute Workflow node. - 5
Fix self-hosted memory issues
Add
--memory 4gto docker run. Enable execution data pruning:N8N_EXECUTIONS_DATA_PRUNE=true. For production: switch from SQLite to PostgreSQL.
Get alerted when n8n goes down
Star n8n on Prismix and get emailed the moment n8n Cloud status changes. Free, no credit card.
Frequently asked questions
Why is n8n not working?
n8n issues: (1) AI agent errors — no tools connected, wrong model name, missing LLM credential; (2) webhook not receiving — workflow not ACTIVE or WEBHOOK_URL not set on self-hosted; (3) timeout — Cloud 5-min limit, set EXECUTIONS_TIMEOUT on self-hosted; (4) memory OOM — increase Docker memory for AI workloads; (5) cloud outage — check prismix.dev/service/n8n.
Is n8n down right now?
Check prismix.dev/service/n8n for live n8n Cloud status. Also n8n-cloud.statuspage.io. Self-hosted n8n issues are server-specific.
n8n AI agent "No tools connected" error — how to fix?
AI Agent node requires at least one Tool to be connected. In the n8n canvas, look for the sub-items handle under the AI Agent node (small "+" area). Connect a Tool node there (Calculator, Code, HTTP Request, SerpAPI, etc.). Without tools, the agent immediately errors. Also add a Chat Memory node for conversation persistence.
n8n webhook receiving but workflow erroring — how to debug?
If webhook receives but workflow errors: (1) go to Executions tab — each execution shows which node errored; (2) click the failed execution to see the full data flow; (3) check the HTTP Method setting on the Webhook node matches what's being sent (GET vs POST); (4) add a Set node after Webhook to log $json to see what data arrived; (5) use the "Execute workflow manually" with sample data to test without activating.
n8n self-hosted vs n8n Cloud — which to use for AI agents?
n8n Cloud pros: easy setup, automatic updates, managed hosting, 30-min timeout on Pro plan. Cons: execution timeout limits, cannot install custom npm packages. Self-hosted pros: no timeout limits, full control, custom nodes, run any npm package. Cons: you manage uptime, Docker memory must be sized for AI workloads (4GB+ for complex agents), webhook URL requires public domain.
