Skip to content

privacy.md

Latest commit

 

History

History
290 lines (180 loc) · 11.6 KB

privacy.md

File metadata and controls

290 lines (180 loc) · 11.6 KB
Raw

Privacy & Data Handling

This page explains how oc-codex-multi-auth handles local data, upstream requests, and debugging artifacts.

Last updated: 2026-07-18

Overview

This plugin prioritizes local control and transparency. It does not ship product telemetry or analytics to third parties. Network traffic is limited to the OpenAI/ChatGPT auth and API endpoints you are actively using, optional Codex instruction/catalog fetches from GitHub, and an optional daily npm version check when auto-update is enabled.

Caution

This plugin is for personal development use with your own ChatGPT Plus/Pro subscription. You are responsible for your prompts, exports, and OpenAI policy compliance.

What We Collect

No first-party telemetry. This plugin does not send usage analytics, crash reports, or account inventories to the package maintainers.

  • No analytics product
  • No usage tracking service
  • No remote logging of prompts to maintainers

Local logs, caches, and config files on your machine are separate; see Data Storage.

Data Storage

All plugin state is stored locally on your machine unless you export it or opt into OS keychain storage.

Account storage (V3)

Item Default path
Global account pool ~/.opencode/oc-codex-multi-auth-accounts.json
Per-project account pool ~/.opencode/projects/<project-key>/oc-codex-multi-auth-accounts.json
Flagged accounts ~/.opencode/oc-codex-multi-auth-flagged-accounts.json

Contents typically include OAuth access/refresh material, account IDs, labels/tags/notes, rate-limit reset metadata, and rotation state. Per-project pools are enabled by default (perProjectAccounts: true).

Codex CLI hydrate (optional source)

On startup the plugin may also read Codex CLI account material under ~/.codex (for example accounts.json) to help bootstrap the local pool. Disable with CODEX_AUTH_SYNC_CODEX_CLI=0. This is local filesystem access only; nothing is uploaded to the package maintainers.

Legacy account filenames (migration only)

Older installs may still have migration sources under ~/.opencode/ or a project tree:

Legacy file Role
openai-codex-accounts.json Pre-rename account pool seed
openai-codex-flagged-accounts.json Pre-rename flagged metadata
openai-codex-blocked-accounts.json Older blocked-account list (migrated into flagged handling)
<project>/.opencode/openai-codex-accounts.json In-repo legacy pool (read for migration; current pools live under ~/.opencode/projects/…)

Current canonical names use the oc-codex-multi-auth-*.json prefix. Do not hand-edit legacy files unless you are recovering an old backup.

Plugin configuration

Item Default path
Plugin config ~/.opencode/openai-codex-auth-config.json

Includes runtime options such as retry profile, rotation strategy, model account pools, TUI preferences, and beginner-safe mode.

OpenCode host files

Item Default path
OpenCode config ~/.config/opencode/opencode.json
OpenCode TUI config ~/.config/opencode/tui.json
OpenCode auth tokens ~/.opencode/auth/openai.json

Optional OS keychain

When CODEX_KEYCHAIN=1 is set, account pools can be stored in the OS credential store (macOS Keychain, Windows Credential Manager, Linux libsecret) under service name oc-codex-multi-auth. JSON files may be renamed with a .migrated-to-keychain.<timestamp> suffix for rollback. Keychain failures fall back to JSON without silently deleting credentials.

TUI quota cache

Item Default path
TUI quota cache $OPENCODE_STATE_DIR when set, otherwise OpenCode's state directory (typically ~/.local/state/opencode/), file oc-codex-multi-auth-tui-quota.json, with a ~/.opencode/ fallback in some builds

Caches recent quota/usage snapshots for prompt status display.

Session recovery storage (host OpenCode)

When sessionRecovery is enabled (default), recoverable session repairs read/write OpenCode's on-disk message/part store:

Item Default path
OpenCode storage root $XDG_DATA_HOME/opencode/storage (macOS/Linux default ~/.local/share/opencode/storage; Windows %APPDATA%/opencode/storage)
Messages …/message/{sessionID}/…
Parts …/part/{messageID}/*.json

Only known structural recovery cases are patched (missing tool results, thinking-block order, thinking-disabled violations). Tokens are not written here.

Catalog and instruction caches

Item Default path
Cache directory ~/.opencode/cache/

May include Codex system instructions, catalog-derived instruction files, ETag/meta files, and the auto-update check cache (update-check-cache.json).

Debug logs

Item Default path
Request logs ~/.opencode/logs/codex-plugin/

Written only when request logging is enabled (ENABLE_PLUGIN_REQUEST_LOGGING=1). Metadata logs omit raw bodies by default; set CODEX_PLUGIN_LOG_BODIES=1 only when you need raw request/response payloads (sensitive: may include prompts and model output).

Backups and exports

Export/import and installer backups may create files under ~/.opencode/backups/ or project-scoped backup directories. Treat exports as credential-bearing data.

Data Transmission

Direct to OpenAI / ChatGPT

API and auth traffic go directly from your machine to OpenAI/ChatGPT endpoints over HTTPS. There is no maintainer proxy.

Service Endpoint
OAuth authorize / token https://auth.openai.com/... (e.g. /oauth/authorize, /oauth/token)
Codex API https://chatgpt.com/backend-api/codex/responses
Usage / quota window https://chatgpt.com/backend-api/wham/usage
Reset-credit listing (codex-reset) https://chatgpt.com/backend-api/wham/rate-limit-reset-credits
Reset-credit redemption (codex-reset) https://chatgpt.com/backend-api/wham/rate-limit-reset-credits/consume

The wham endpoints are the same undocumented endpoints the official Codex clients use for the /usage screen and reset-credit redemption. They are reached only from quota-facing surfaces — codex-limits, codex-reset, the "Check quotas" action in the interactive auth dashboard, and the TUI quota status refresh. The model request path does not call them.

What gets sent on a normal model request

When you use the plugin, a request can include:

  • Your prompts and conversation history (as supplied by OpenCode)
  • OAuth access token (for authentication)
  • ChatGPT account/workspace identifiers used for routing
  • Model selection, reasoning effort, verbosity, and related options
  • Client identity headers: originator and a product User-Agent (Codex CLI style or host/opencode style, depending on model and config). These are sent; they are not suppressed by default.
  • For GPT-5.6: responses-lite markers such as x-openai-internal-codex-responses-lite

This is analogous to what official Codex-style clients send when talking to the same backend. Exact fields vary by model family and transform mode.

What does not get sent to maintainers

  • No automatic upload of account lists, tokens, or logs to the plugin authors
  • No remote analytics endpoint for this package

Optional npm auto-update check

When autoUpdate is enabled (default true; disable with autoUpdate: false or CODEX_AUTH_AUTO_UPDATE=0), the plugin may query the public npm registry (registry.npmjs.org/oc-codex-multi-auth/latest) about once per day to detect a newer version, cache the result under ~/.opencode/cache/, and clear the OpenCode-managed plugin cache so a restart can pick up the update. That request does not send your ChatGPT tokens or prompts.

Third-Party Services

GitHub API

The plugin may fetch Codex instructions / model catalog material from GitHub (for example release metadata under openai/codex) with local caching and ETag reuse. Requests are ordinary HTTPS GETs without your ChatGPT credentials.

OpenAI Services

All auth and inference go through OpenAI/ChatGPT as listed above. See OpenAI Privacy Policy for how OpenAI handles data.

Your Data Rights

You have complete control over local data:

Delete OAuth Tokens

opencode auth logout
# Or manually:
rm ~/.opencode/auth/openai.json

Delete Account Pools and Flagged State

rm ~/.opencode/oc-codex-multi-auth-accounts.json
rm ~/.opencode/oc-codex-multi-auth-flagged-accounts.json
rm -rf ~/.opencode/projects/

Also remove any project-scoped account files and keychain entries if you migrated with CODEX_KEYCHAIN=1 (see codex-keychain).

Delete Plugin Config, Caches, Logs, Quota Cache

rm ~/.opencode/openai-codex-auth-config.json
rm -rf ~/.opencode/cache/
rm -rf ~/.opencode/logs/codex-plugin/
rm -f ~/.opencode/oc-codex-multi-auth-tui-quota.json

Revoke OAuth Access

  1. Visit ChatGPT Settings → Authorized Apps
  2. Find the app entry used for login (OpenCode / Codex-related)
  3. Click Revoke

This invalidates access tokens for that authorization.

Security Measures

Token Protection

  • Tokens stay local except when sent to OpenAI/ChatGPT for authentication and API calls
  • Auth and account files should remain user-readable only where the OS allows
  • Diagnostic tools redact tokens and sensitive identifiers by default
  • Expired tokens are refreshed automatically; refresh is queued to avoid races

PKCE Flow

The plugin uses PKCE for the browser OAuth flow (same class of flow used by official Codex CLI login).

HTTPS Encryption

OAuth, token refresh, and API requests use HTTPS.

Email Masking in Account Displays

Account emails can appear in screenshots or shared TUI sessions. To reduce exposure:

  • Set a non-identifying label with codex-label (labels are preferred over emails)
  • Enable maskEmail in ~/.opencode/openai-codex-auth-config.json (or CODEX_TUI_MASK_EMAIL=1) so remaining emails render as forms like us***@example.com
  • Raw emails appear in --includeSensitive / includeSensitive JSON output only when you opt in

Compliance

OpenAI Policies

When using this plugin, you are subject to:

Your responsibility: ensure usage complies with OpenAI's policies and your subscription terms.

Local Data Control

This plugin:

  • Does not operate a maintainer-side personal data processing service
  • Stores operational state locally under your control
  • Provides deletion steps for local files and OAuth revocation

Data sent to OpenAI remains subject to OpenAI's practices.

Transparency

Open Source

Source: https://github.com/ndycode/oc-codex-multi-auth

You can review request shaping, storage, and logging behavior in the repository.

No Hidden Telemetry Product

There is no separate analytics backend for this package. Documented network calls are OAuth/API, optional GitHub catalog fetches, and optional npm version checks.

Questions?

Back to: Documentation Home | Getting Started | Tools and CLI