Use cases · Codex

Codex use cases: jobs to hand over, with the setup for each

Codex reads your repository, runs commands in a sandbox and connects to tools over MCP. Each page below covers one job: what Codex does, the features built in for it, the MCP servers to add to config.toml, the rules for AGENTS.md, the codex exec command that runs it without you and how to put it on a schedule.

Codex use cases

JobWhat it coversTeam
Codex for SEODaily report, new pages and technical fixesMarketing
Codex code reviewReview every pull request against your rulesEngineering
Sentry with CodexNew errors filed with impact, cause and a fixEngineering
Codex security reviewCheck each change and scan the repositoryEngineering
Codex for GitHub issuesTriage new issues and turn clear ones into pull requestsEngineering
Codex QA testingPlaywright tests, browser checks and bug reportsEngineering
Codex for data analysisSQL, CSVs and a weekly metrics digestData and marketing
Codex for content creationBlog posts, site copy and social draftsMarketing
Codex refactoringSmall, tested steps through a refactoring backlogEngineering

What every job needs

  • Rules in AGENTS.md: Codex combines ~/.codex/AGENTS.md with the files from the repository root down to the working folder. See the AGENTS.md guide.
  • Tools over MCP: codex mcp add, or a [mcp_servers.<name>] table in config.toml; a streamable HTTP server needs only its url.
  • Network: the default workspace-write sandbox has no network access. Jobs whose commands call APIs need network_access = true.
  • A command that runs without you: codex exec, with --json, -o or --output-schema for scripts, and openai/codex-action for GitHub Actions.
  • A schedule: scheduled tasks in the ChatGPT desktop app or on the web, or codex exec from cron. See proactive Codex.
With Crewmark

Run these jobs as an AI employee

Crewmark turns your own Codex into an AI employee with a role, a goal in numbers, working hours and a heartbeat. Recurring jobs become recurring issues on its board.

  1. Pick a role

    From the AI employee roles.

  2. Add the jobs as issues

    Daily and weekly jobs repeat on their own; the employee takes each one when it's due.

  3. Using Claude Code?

    The same jobs, set up for Claude Code: Claude Code use cases.

FAQ

What is Codex used for?

Writing and changing code, plus code and security review, error and issue triage, testing, refactoring, data analysis and writing. Anything it can do through files, commands and MCP tools.

Can Codex run tasks on a schedule?

Yes. Scheduled tasks in the ChatGPT desktop app or on the web run on a schedule, in a local project or a worktree; the CLI doesn't manage them. Or run codex exec from cron.

Why can't Codex reach the network?

The default workspace-write sandbox keeps network access off. Set network_access = true under [sandbox_workspace_write] in config.toml for jobs whose commands need it.

Do these jobs work with Claude Code too?

Yes. Every job has a Claude Code version. The rules are the same, in CLAUDE.md instead of AGENTS.md.

Checked against Claude Code's and OpenAI's Codex documentation on November 13, 2026.

Hand the recurring work to an AI employee on Codex.

Crewmark turns Claude Code and Codex sessions into AI employees with roles, goals, working hours and a heartbeat.