What version of the Codex App are you using (From “About Codex” dialog)?
26.601.21317
What subscription do you have?
Pro
What platform is your computer?
Darwin 24.5.0 arm64 arm
What issue are you seeing?
A Codex desktop automation that runs a networked command on a cadence can fail with host resolution / outbound connectivity errors for external APIs, while those same hosts and endpoints work from the interactive Codex shell and from Postman on the same machine.
Observed pattern:
- Background automation runs fail on external network calls
- The same hosts are reachable from the interactive Codex shell
- The same requests work in Postman on the same machine
Examples we encountered during debugging included failures against SEC-related hosts and other third-party APIs such as Marketaux, which makes this look broader than a single provider-specific outage or request-shape bug.
What steps can reproduce the bug?
- Create or run a Codex desktop automation that executes a networked command on a cadence.
- Have that command call one or more external APIs during execution.
- Observe that the background automation can fail with host lookup / outbound network errors (ie,
dial tcp: lookup api.abc.com: no such host).
- In the same Codex app session, test the same hosts manually from the interactive shell with tools like
dig and curl.
- Also test the same request from Postman on the same machine.
- The interactive shell and Postman succeed, while the background automation path fails.
One concrete reproduction we hit was a bounded Go backfill command that made HTTPS requests to external filing/news providers and failed only in the automation/background context.
What is the expected behavior?
Background automations should have network resolution and outbound HTTPS behavior that is consistent with the interactive Codex shell on the same machine, or the product should clearly surface when the background automation runtime is operating under different DNS/network constraints.
Additional information
Notes:
- This was on the Codex desktop app on macOS.
- The issue appears environment-specific to the automation/background runner rather than a single external provider.
- During debugging, equivalent requests succeeded from Postman and from the interactive Codex shell on the same machine.
- We saw this pattern with more than one host, including SEC-related endpoints and Marketaux.
What version of the Codex App are you using (From “About Codex” dialog)?
26.601.21317
What subscription do you have?
Pro
What platform is your computer?
Darwin 24.5.0 arm64 arm
What issue are you seeing?
A Codex desktop automation that runs a networked command on a cadence can fail with host resolution / outbound connectivity errors for external APIs, while those same hosts and endpoints work from the interactive Codex shell and from Postman on the same machine.
Observed pattern:
Examples we encountered during debugging included failures against SEC-related hosts and other third-party APIs such as Marketaux, which makes this look broader than a single provider-specific outage or request-shape bug.
What steps can reproduce the bug?
dial tcp: lookup api.abc.com: no such host).digandcurl.One concrete reproduction we hit was a bounded Go backfill command that made HTTPS requests to external filing/news providers and failed only in the automation/background context.
What is the expected behavior?
Background automations should have network resolution and outbound HTTPS behavior that is consistent with the interactive Codex shell on the same machine, or the product should clearly surface when the background automation runtime is operating under different DNS/network constraints.
Additional information
Notes: