Privacy Policy

Scope

Filament Labs develops Codex Process Jobs (CPJ). This policy covers the local plugin, its project website, and messages sent to Filament Labs about CPJ.

Data processed and purposes

CPJ runs commands on your macOS or Linux machine. It stores job names, command arguments, working directories, task identifiers, process identifiers, timestamps, exit status, notification state, and bounded standard output and standard error logs. These records support detached execution, progress checks, result retrieval, completion delivery, cancellation, and reruns.

Command arguments and process output can contain personal or confidential information. CPJ does not automatically remove secrets from those fields. Avoid placing secrets in command arguments or printing them in process output.

CPJ does not operate a hosted job service or send job records, process output, or plugin telemetry to Filament Labs. CPJ does not save a copy of the launching process's environment. A launched command inherits its environment and may write information from it into its own output or files.

Local storage, retention, and controls

CPJ normally stores its job records and logs in the process-jobs directory under your Codex home. A custom Codex home changes that location. Output logs are size-limited and retain recent output when the limit is reached. A size limit is not a time-based deletion policy.

Saved job records and logs have no automatic age-based expiration. They can remain after a command ends or the plugin is removed. You can inspect the saved files and manually delete records and logs for completed jobs. Do not remove files that an active job or notification relay still uses. Files created by your commands are separate from CPJ's records and follow the command's own storage behavior.

Copies in device backups, exported results, or conversations can remain after local deletion. Use the controls for those systems to manage their copies.

Completion delivery and local hooks

On supported local tasks, CPJ reads local Codex lifecycle and task-relationship metadata to identify the owning task and determine whether it is idle. Its approved prompt hook also processes bounded prompt text to route job requests and present unread results. Local plugin hooks do not run under cloud orchestration.

Automatic model-facing completion notices contain validated job identifiers, terminal status, exit codes, and fixed instructions. They do not include process output or command arguments. If enabled, a system notification can display the job label. The agent can retrieve saved output in a later tool call.

Recipients and host services

An agent can inspect job metadata and selected saved output. OpenAI may process those results under your account settings and its privacy policy. OpenAI controls retention of its service data and conversations. Use your OpenAI account's data controls to manage that data.

Commands launched through CPJ may contact other services or write additional files when authorized. Those services and commands have their own data practices. CPJ does not grant itself permissions or change your host's access policy.

Website services and analytics

Cloudflare hosts and delivers this website. It receives network information to serve and protect the site. This includes your IP address and the requested URL. The project pages also use Cloudflare Web Analytics to measure aggregate traffic and page performance. Analytics dimensions include visited pages, referring sites, country, browser, operating system, and device type. Performance measurements include page-load timing.

Cloudflare states that Web Analytics does not use cookies, local storage, or browser fingerprinting to track visitors. Its Web Analytics documentation states that unsampled beacon data is kept for seven days, followed by sampled data for longer-term reporting. The dashboard provides access to the previous six months. This reporting window is not a promise that all Cloudflare service data is deleted after six months. Cloudflare's privacy policy describes its retention practices and privacy rights.

You can use browser controls or a content blocker to prevent the analytics script from loading. This does not prevent Cloudflare from processing the requests needed to deliver the site. Filament Labs does not use advertising trackers or a website contact form.

The pages load fonts from Google Fonts. Your browser sends requests to Google to retrieve those fonts. Google receives the network information needed to answer those requests. Its processing and retention are governed by Google's privacy policy. You can block external font requests with browser controls; the pages can use fallback fonts.

Support messages and privacy requests

If you email Filament Labs, we receive your email address, any name you provide, and the contents and attachments you send. We use these details to answer your request and investigate issues. Filament Labs and the email providers that deliver and store the correspondence process these messages.

Support correspondence remains in the email system until manually deleted, including in response to a deletion request. This policy does not promise an automatic deletion deadline. You can request access, correction, or deletion of correspondence you sent by emailing [email protected]. We may need information to locate the message and verify that the request is yours. The email provider's backup and deletion practices can affect when removed messages cease to exist in its systems.

Public GitHub issues and comments are visible to other people. GitHub processes those submissions under its privacy statement. Use GitHub's controls to manage your submissions. Do not send or post credentials, private logs, or sensitive process output. Send suspected vulnerabilities privately through the support contact.

Contact and changes

Send privacy questions to [email protected]. We publish policy changes on this page and update the date above.