Codex articles
Configuring and using OpenAI's Codex CLI, VS Code extension and cloud agent
-
Allow only some tools from an MCP server in Codex: enabled_tools, timeouts, and approval modes
Restrict which tools an MCP server exposes to Codex with enabled_tools and disabled_tools, tune startup_timeout_sec and tool_timeout_sec, and require approval before tools run.
-
Control Codex web search with the web_search setting: disabled, cached, indexed, and live
The four values of Codex CLI's web_search setting and their defaults, the --search flag for one-off use, why indexed reduces prompt injection risk, and how to restrict domains.
-
Get a desktop or Slack notification when Codex finishes a turn: notify and tui.notifications
Run a program when Codex CLI completes a turn with the notify setting, the JSON it receives, and how tui.notifications filters terminal notifications, with macOS/Linux examples.
-
Hide or expand reasoning in Codex output: hide_agent_reasoning and model_reasoning_summary
The three Codex CLI settings that control reasoning display: hide_agent_reasoning, show_raw_agent_reasoning, and model_reasoning_summary, and how they differ from reasoning effort.
-
Keep API keys out of the commands Codex runs: the shell_environment_policy setting
Control which environment variables reach the commands Codex CLI runs with shell_environment_policy: inherit values, default handling of KEY/SECRET/TOKEN names, filters, and set.
-
Let Codex write outside the workspace in workspace-write mode: writable_roots and the /tmp rules
What Codex's workspace-write sandbox can write to by default (cwd, /tmp, $TMPDIR), how to add paths with writable_roots, how to exclude temp directories, and why .git is read-only.
-
Manage Codex's "trust this folder?" prompt in config.toml with the projects trust_level setting
How Codex records project trust under [projects."<absolute path>"] trust_level, what it ignores for untrusted projects (.codex/config.toml, hooks, rules), and its precedence.
-
Open file references from Codex output directly in VS Code or Cursor with file_opener
How the file_opener setting in Codex CLI turns cited file paths into clickable editor links: the values (vscode, vscode-insiders, windsurf, cursor, none) and terminal requirements.
-
Stop Codex saving conversation history to disk: history.persistence and where credentials are stored
Keep Codex CLI from writing conversation history with history.persistence = "none", cap it with max_bytes, and choose where login credentials live with cli_auth_credentials_store.
-
Why Codex reads only part of AGENTS.md: project_doc_max_bytes and fallback file names
How Codex discovers AGENTS.md (search order, AGENTS.override.md, the 32 KiB default cap, fallback names) and how project_doc_max_bytes and project_doc_fallback_filenames change it.
-
Codex CLI approval modes versus sandbox modes: safe combinations for autonomous runs
What Codex CLI's approval policy and sandbox mode each control, what --full-auto really does, how to pin them in config.toml, and which pair fits research, daily work and CI.
-
Codex vs Claude Code: pricing, permission models, config files and when to use which
OpenAI's Codex and Anthropic's Claude Code compared on what changes slowly: pricing, permission and sandbox design, AGENTS.md versus CLAUDE.md, cloud execution and CI.
-
Switching models, reasoning effort and profiles in Codex CLI's config.toml
The settings you actually touch in ~/.codex/config.toml: model and reasoning effort, task-specific profiles with --profile, and how options, profiles and defaults take precedence.