Codex

Automation and cloud runs

Schedule project checks, retain reusable workflows and distinguish local from cloud execution.

#Automation

With recurring triggers, Codex can inspect a project, find issues, handle bounded tasks and prepare results for review instead of waiting for each manual instruction.

#Turn session retrospectives into reusable workflows

A practical starting point is a weekly review of accessible execution records that extracts lessons and working preferences:

Weekly session retrospective

Search and review the last week's Codex conversations and execution logs.
Maintain a session retrospective and personal working-style document.

Requirements:
1. Prefer available history-search tools. For logs, search metadata and relevant
   excerpts rather than loading entire large session files.
2. Do not reproduce raw logs, personal information, keys or long transcripts.
3. Summarize what caused problems, the correct approach and reusable scenarios.
4. Record UI, product, interaction, content and workflow preferences.
5. Produce reusable rules, merge duplicates and cite dates and task types.
6. Assess which rules belong in project or user AGENTS.md files.

Recurring runs can accumulate a useful record if the workflow can access and update it. Persisted instructions and available memory features determine what future sessions can reuse; running a prompt alone does not guarantee cross-session memory.

#Local versus cloud execution

Three common execution locations are:

Mode	Location	Source	Suitable use
Local	Your computer	Local project directory	Everyday work and local private files
Worktree	Your computer	Isolated Git checkout	Experiments without modifying the main working directory
Cloud	Managed cloud environment	Connected repository	Background and parallel tasks

#How cloud execution works

Cloud tasks work in a hosted environment: obtain repository code, inspect and modify files, run checks and prepare a diff or pull request. They do not require the model task to occupy your local machine.

Suitable: bug fixes, features, documentation, code review and issue work.

Unsuitable without additional setup: private local files, unpushed projects and high-risk production operations.

Dimension	Local	Cloud
Code location	Your computer	Connected repository
Execution	Your computer	Cloud environment
Local resource use	Yes	Minimal for the hosted task
GitHub requirement	Not always	Usually a connected repository
Background / parallel work	Depends on setup	Designed for isolated tasks
Permissions	Local file and execution access	Repository, environment and network policy
Starting point	Start here	Use after understanding repository connections

#Run a cloud task

Push the repository

Ensure the project is pushed and the official cloud workspace can access it.

Open the official cloud interface

Use the current official Codex web/cloud entry point and its required account authentication. This is a separate official service, not a Passion8 login path.

Select the repository and task

Choose the repository and describe the desired outcome.

Wait for the result

Review the resulting diff or pull request and the checks performed.

Synchronize local code

Before further local development, update your checkout to avoid working against stale code:

git pull
# Or use Codex apply for the relevant remote task's diff.

Official cloud execution does not automatically inherit a local Passion8 base_url. It has its own account, workspace and environment requirements.

For the documented Passion8 setup, use local CLI or supported local desktop workflows. See CLI configuration and Desktop app.

#Continue reading

CLI configurationPassion8 authentication and configuration files.Desktop appLocal visual workflows and project management.

PreviousSubagents and worktrees NextNon-interactive execution, CI and SDKs

Support

Need help?

For setup, billing, or model issues, email us. Check the status page for uptime.

Support Email[email protected]Service Statusstatus.passion8.cc

WeChat / QQ support is available at the bottom right.