Codex CLI 0.162.0 won't start: `account/read failed during TUI bootstrap: account/read failed: workspace routing discovery failed (code -32603)`

Context

Source(s): https://community.openai.com/t/codex-cli-0-162-0-fails-during-tui-bootstrap-workspace-routing-discovery-failed-behind-a-local-proxy/1404666 Signal: 713 views in 9 days; the original poster marked the fix as the solution (October 2026). Status: reported working by source, not reproduced by us.

Problem

After a network interruption and a computer restart, codex exits on every launch during TUI bootstrap with the error above, before any model request. codex login status still says "Logged in using ChatGPT", and moving config.toml out of the way does not change anything. RUST_LOG=debug only prints the same line.

Error (redacted / normalized)

Error: account/read failed during TUI bootstrap: account/read failed: workspace routing discovery failed (code -32603)

Applies when

Codex CLI 0.162.0 (Linux x86_64 musl build from the official GitHub Releases), Linux Mint, ChatGPT login, traffic going through a local proxy app.

Confirm it is this problem

1. codex --version and codex app-server daemon version → expect: the two versions differ (in the thread the CLI was 0.162.0 while the background app-server was still 0.161.0).
2. codex doctor --summary --no-color → expect: local databases and the install look healthy and the WebSocket check succeeds (HTTP 101), so the problem is not the install itself.

Not this when / boundary

The versions match, the daemon restart does not help, and no proxy is involved; then the cause is something else (our inference).

Root cause

The original poster found the CLI and the persistent background app-server on different versions, and also a wrong proxy port. Which of the two caused the error was not confirmed; restarting the app-server is what fixed it. Root cause not confirmed.

Fix

1. Run codex app-server daemon restart, then start codex again. → expect: the original poster said Codex then launched normally, without deleting config, credentials or conversation history.
2. If you use a local proxy, check that Codex points at the proxy's HTTP port, not its SOCKS port (the original poster had them swapped; a curl -I --proxy ... test returned 407 Proxy Authentication Required until it was corrected). → expect: the 407 goes away. In the thread this alone did not fix Codex until the daemon was restarted.

How to verify

codex opens the interactive terminal UI, and codex app-server daemon version matches codex --version.

What did not work

Moving config.toml aside, RUST_LOG=debug (no extra detail), fixing the proxy port without restarting the daemon (all from the original poster).

Risks

Restarting the daemon may interrupt active Codex sessions (the original poster). Finish or stop running tasks first.

Verified scope

Codex CLI 0.162.0 with a background app-server from an older release, October 2026.

Becomes obsolete when

The CLI restarts or rejects an outdated app-server on its own after an upgrade.