Codex Auth Fallback

Multi-provider auth setup for OpenClaw with automatic failover between Anthropic and multiple OpenAI Codex OAuth sessions.

Overview

OpenClaw supports multiple auth profiles per provider. When one profile hits a rate limit, the platform can fail over to another. This skill covers:

1. Adding Codex OAuth profiles via device-flow login
2. Configuring openclaw.json for provider fallback order
3. Setting up auth-profiles.json with multiple profiles
4. Deploying a cron job to auto-switch models on cooldown

Prerequisites

• OpenClaw instance running
• codex CLI installed (npm i -g @openai/codex) — this also ensures node is available
• One or more OpenAI accounts with Codex access

Security & Safety

What this skill accesses:

File	Access	Purpose
~/.codex/auth.json	Read + Temporary Write	Temporarily cleared to force a fresh device-flow login, then restored from backup. Original tokens are never deleted — a timestamped backup is created first.
~/.openclaw/agents/main/agent/auth-profiles.json	Read + Write	Imported OAuth tokens (access + refresh) are written here. A timestamped backup is created before any modification.

Important safety notes:

• Tokens stay local. No tokens are sent to any external endpoint. The script reads tokens from the local Codex CLI auth file and writes them to the local OpenClaw auth-profiles file.
• Backups are always created. Both files are backed up with timestamps before any modification. If login fails or the script is interrupted, a trap handler restores the original Codex CLI auth automatically.
• Interactive confirmation. The script prompts for confirmation before clearing the Codex CLI auth file, so you can abort if needed.
• No elevated privileges. The script runs as your user and does not require sudo or any special permissions.
• Back up manually first. Despite the automatic backups, it is recommended to manually back up ~/.codex/auth.json and your OpenClaw configs before running, especially on first use.
• Test with a non-production account. For initial testing, consider using a throwaway or non-production OpenAI account.

Step 1: Add Codex OAuth Profiles

Run the bundled script for each OpenAI account:

./scripts/codex-add-profile.sh <profile-name>

The script:

1. Backs up ~/.codex/auth.json and auth-profiles.json
2. Clears Codex CLI auth to force fresh device-flow login
3. Runs codex auth login (opens browser for OAuth)
4. Extracts tokens and imports them into OpenClaw's auth-profiles.json
5. Restores the original Codex CLI auth

Repeat for each account. Profile names should be short identifiers (e.g., the OpenAI username).

Step 2: Configure openclaw.json

Add auth profile declarations and fallback model config. See references/config-templates.md for the exact JSON blocks to add to openclaw.json.

Key sections:

• auth.profiles — Declare each profile with provider and mode
• auth.order — Set failover priority per provider
• agents.defaults.model — Set primary model + fallbacks

Step 3: Auth Profiles JSON Structure

OpenClaw stores live tokens in agents/main/agent/auth-profiles.json. See references/config-templates.md for the schema.

Each Codex profile contains:

• type: "oauth"
• provider: "openai-codex"
• access: JWT access token (auto-populated by the add-profile script)
• refresh: Refresh token (auto-populated)
• expires: Token expiry in ms (parsed from JWT)
• accountId: OpenAI account ID (parsed from JWT)

The order object controls which profile is tried first per provider. The usageStats object tracks rate limits and cooldowns automatically.

Step 4: Model Cooldown Auto-Switch Cron (Optional)

This step is entirely optional. The auth profiles from Steps 1-3 work on their own with OpenClaw's built-in failover. This cron job adds automatic model switching, which means your active model may change without manual intervention. Only enable it if you understand and want this behavior.

Deploy a cron job that checks cooldown state every 10 minutes and switches the active model. See references/config-templates.md for the full cron job definition.

The cron job:

1. Runs openclaw models status to check cooldown state
2. Picks the best available model (priority: opus > codex profiles in order)
3. Updates the session model override if needed
4. Logs state to a local memory file; only notifies on change

Before enabling:

• Test manually first: run openclaw models status to verify your profiles are working
• Review the cron job template in references/config-templates.md — the job only runs local commands and writes to a local state file
• The job runs in an isolated session and does not affect your main chat unless a model switch occurs

Add the job to cron/jobs.json using the template in the references.

File Layout

codex-auth-fallback/
├── SKILL.md                    # This file
├── scripts/
│   └── codex-add-profile.sh    # Device-flow profile importer
└── references/
    └── config-templates.md     # openclaw.json, auth-profiles, cron templates