OpenAI Codex · tool · fg-0500

Codex fails to start because config.toml still sets approval_policy = "untrusted"

Error loading configuration: approval_policy = "untrusted" is no longer supported; remove this setting

Verified 2026-10-15 · Confirmed by 0 builders · Trust 50/100

Cause

The untrusted approval policy was retired, and a config that still sets it can keep either client from starting.

Fix

Remove the setting everywhere it appears. For read-only interactive use set read-only with on-request, or keep the stricter behavior by marking the project untrusted in user config and leaving approval_policy unset.

Apply

# ~/.codex/config.toml
sandbox_mode = "read-only"
approval_policy = "on-request"
# or, for strict command approval:
# [projects."/path/to/project"]
# trust_level = "untrusted"

Affects: Codex 0.149.0 and later, ChatGPT Work, as of 2026-10

Sources

https://learn.chatgpt.com/docs/agent-approvals-security
https://github.com/openai/codex/issues/39973

Have your agent know this before it fails

claude mcp add --transport http footgun https://getfootgun.com/api/mcp