Skip to main content
Several Zespan features call a URL you configure: an LLM connection’s base URL, a simulation or dataset HTTP target, a webhook, or your SSO provider’s discovery document. Zespan only sends these requests to public internet addresses. A request is refused if its host resolves to a private, loopback, link-local, or cloud-metadata address.

What the policy covers

  • LLM connection base URLs, including model listing, Test, and the Playground
  • Simulation and dataset HTTP targets
  • Webhooks: alert webhooks (generic, Slack, Discord), prompt webhooks (generic, Slack), Jira, and Freshservice
  • SSO discovery and key-set URLs
  • Agent card fetches
Hostnames are resolved and checked again on every request, not only when you save the URL. A hostname that later starts resolving to a private address is blocked from then on.

Redirects are not followed

If a target answers with a 3xx redirect, Zespan does not follow it. The simulation run, dataset run, or webhook delivery fails instead. Point the target at its final URL.

Error messages

Running a private LLM gateway on a self-hosted deployment? Operators can allow specific private endpoints for LLM connections only. See LLM Connections → Private and self-hosted endpoints.

Self-hosted deployments

Requests covered by this policy connect directly and don’t use the HTTP_PROXY or HTTPS_PROXY environment variables. If your network requires all egress to go through a proxy, enforce an equivalent policy at the proxy.

Next steps