This page explains how oc-codex-multi-auth handles local data, upstream requests, and debugging artifacts.
This plugin prioritizes user privacy and data security. We believe in transparency about data handling and giving you full control over your information.
Nothing. This plugin does not collect, store, or transmit usage data to third parties.
- ❌ No telemetry
- ❌ No analytics
- ❌ No usage tracking
- ❌ No personal information collection
All data is stored locally on your machine:
- Location:
~/.opencode/auth/openai.json - Contents: Access tokens, refresh tokens, expiration timestamps
- Managed by: OpenCode's credential management system
- Security: File permissions restrict access to your user account
- Location:
~/.opencode/cache/ - Contents:
codex-instructions.txt- Codex system instructions (fetched from GitHub)codex-instructions-meta.json- ETag and timestamp metadata
- Purpose: Reduce GitHub API calls and improve performance
- TTL: 15 minutes (automatically refreshes when stale)
- Location:
~/.opencode/logs/codex-plugin/ - Contents: Request/response metadata logs (only when
ENABLE_PLUGIN_REQUEST_LOGGING=1is set) - Includes:
- Request metadata (model, flags, response status, timing)
- Raw request/response payloads only when
CODEX_PLUGIN_LOG_BODIES=1is also set - Timestamps
- Configuration used
⚠️ Warning: Logs may contain your prompts and model responses - handle with care
All API requests go directly from your machine to OpenAI's servers:
- ✅ No intermediary proxies
- ✅ No third-party data collection
- ✅ HTTPS encrypted communication
- ✅ OAuth-secured authentication
When you use the plugin, the following is transmitted to OpenAI:
- Your prompts and conversation history
- OAuth access token (for authentication)
- ChatGPT account ID (from token JWT)
- Configuration options (reasoning effort, verbosity, etc.)
- Model selection
Note: This is identical to what the official OpenAI Codex CLI sends.
- ❌ Your filesystem contents (unless explicitly requested via tools)
- ❌ Personal information beyond what's in your prompts
- ❌ Usage statistics or analytics
- ❌ Plugin version or system information
The plugin fetches Codex instructions from GitHub:
- URL:
https://api.github.com/repos/openai/codex/releases/latest - Purpose: Get latest Codex system instructions
- Frequency: Once per 15 minutes (cached with ETag)
- Data sent: HTTP GET request (no personal data)
- Rate limiting: 60 requests/hour (unauthenticated)
All interactions with OpenAI go through:
- OAuth:
https://chatgpt.com/oauth - API:
https://chatgpt.com/backend-api/conversation
See OpenAI Privacy Policy for how OpenAI handles data.
You have complete control over your data:
opencode auth logout
# Or manually:
rm ~/.opencode/auth/openai.jsonrm -rf ~/.opencode/cache/rm -rf ~/.opencode/logs/codex-plugin/- Visit ChatGPT Settings → Authorized Apps
- Find "OpenCode" or "Codex CLI"
- Click "Revoke"
This immediately invalidates all access tokens.
- Local storage only: Tokens never leave your machine except when sent to OpenAI for authentication
- File permissions: Auth files are readable only by your user account
- No logging: OAuth tokens are never written to debug logs
- Automatic refresh: Expired tokens are refreshed automatically
The plugin uses PKCE (Proof Key for Code Exchange) for OAuth:
- Prevents authorization code interception attacks
- Industry-standard security for OAuth 2.0
- Same method used by OpenAI's official Codex CLI
All network communication uses HTTPS:
- OAuth authorization: Encrypted
- API requests: Encrypted
- Token refresh: Encrypted
When using this plugin, you are subject to:
Your responsibility: Ensure your usage complies with OpenAI's policies.
This plugin:
- ✅ Does not collect personal data
- ✅ Does not process data on behalf of third parties
- ✅ Stores data locally under your control
- ✅ Provides clear data deletion mechanisms
However, data sent to OpenAI is subject to OpenAI's privacy practices.
The entire plugin source code is available at:
You can:
- Review all code
- Audit data handling
- Verify no hidden telemetry
- Inspect network requests
No Hidden Behavior
- No obfuscated code
- No minified dependencies
- All network requests are documented
- Debug logging shows exactly what's sent to APIs
For privacy-related questions:
-
Plugin-specific: GitHub Issues
-
OpenAI data handling: OpenAI Support
-
Security concerns: See SECURITY.md
Last Updated: 2026-03-11
Back to: Documentation Home | Getting Started