n8n Fails to Connect to External APIs via Corporate Proxy

Below is a summary of some troubleshooting steps that may help resolve issues with n8n not using your corporate proxy:

---

Check That n8n Inherits the Proxy Settings

Even though you’ve set the proxy in your global shell, make sure that the environment variables are actually available in the context from which n8n is running. For example, if running n8n in a Docker container, verify that you pass the proxy settings explicitly with the –env flags:

BASHCopied
docker run -e HTTP_PROXY="http://xx.xx.com:xxxx" \
           -e HTTPS_PROXY="http://xx.xx.com:xxxx" \
           -e NODE_TLS_REJECT_UNAUTHORIZED=0 \
           -p 5678:5678 \
           --name n8n \
           docker.n8n.io/n8nio/n8n

You can use an interactive shell inside the container to confirm the variables:

BASHCopied
docker exec -it n8n env | grep -i proxy

---

Verify Proxy Usage in Nodes

Some n8n nodes (for example, the HTTP Request node) support a manual proxy setting in their parameters. Since your OpenAI integration appears to be calling an external API via the built-in integration (or using a built-in fetch mechanism), it may not automatically honor the OS‑level environment variables.

• If you’ve already tried manually adding the proxy URL in the HTTP Request node’s “Proxy” field and still got errors like EHOSTUNREACH, that suggests the underlying HTTP library isn’t using the proxy settings as expected.

• For nodes that don’t have a dedicated proxy option (such as some integrations like the Azure OpenAI or other LLM integrations), these environment variables may not be picked up automatically by the underlying libraries.

---

Consider Using a Proxy Agent

Because many HTTP libraries (e.g. node‑fetch) do not automatically honor the HTTP_PROXY/HTTPS_PROXY environment variables, you may need to inject a proxy agent so that outbound API calls route through your corporate proxy.

A common workaround is to use an HTTP Request node with its proxy option when available. When that is not possible (such as with built‑in integrations that don’t offer a proxy field), you might have to modify or wrap the outgoing request (for example, using a custom node or an HTTP client library that supports proxy agents).

For example, if you decide to create a custom node or script, you could employ the “global-agent” package:

JSCopied
// At the very start of your application
require('global-agent/bootstrap');

Then, setting the environment variable works as expected and all HTTP calls using the built‑in node‑fetch library will respect your corporate proxy.

---

Other Considerations

• Double‑check that your corporate proxy does not require additional authentication or special CONNECT tunneling settings for HTTPS.

• Confirm that the exact proxy host and port are accessible from the n8n runtime environment (container or host). Although curl and wget work, n8n might be using different methods, so verifying connectivity from within n8n is key.

• In environments where n8n connects to multiple services (like in production workflows versus testing), remember that some integrations (for example, the ones managing webhooks) have their own network asset configuration. This issue is specific to outbound HTTP calls; make sure that any additional internal network configuration (such as firewall rules or corporate proxy ACLs) isn’t interfering.

---

In Summary

Because n8n’s built‑in nodes (especially those calling external APIs like OpenAI) sometimes do not use proxy settings automatically, you might need to:

1. Ensure the environment variables are available in the running environment.
2. For nodes that support it (like HTTP Request), explicitly set the proxy in the node’s options.
3. For built‑in integrations that lack a proxy field, consider using a custom solution or a workaround (e.g. employing a proxy agent library) so that outbound requests are routed correctly.

Review your deployment method and ensure that these settings are being applied. If these steps still don’t resolve the issue, additional debugging (for example, capturing the raw HTTP requests from within n8n) might be necessary.

Hopefully, these insights help you pinpoint why n8n is bypassing your corporate proxy. Let me know if you need further assistance with your workflow configuration.