Generated banner for Migrate To Codex
Generated logo for Migrate To Codex

Migrate To Codex

OpenAIOfficialPublished by OpenAI

  • Skill
  • Workflow
  • Automation and integrations
  • Apache-2.0
  • Low risk
  • Has scripts

From the author

Migrate supported instruction files, skills, agents, and MCP config into Codex project and global files.
— the description by OpenAI

AutoGineer summary

Written by AutoGineer from the skill’s own files, not by the author

The migrate-to-codex skill provides a structured, automated procedure to migrate existing Claude Code setups into the Codex ecosystem. It handles the conversion of instruction files, slash commands, skills, agents, and MCP server configurations. The tool follows a rigorous process involving scanning, planning, dry-running, and validating the migration to ensure artifacts are correctly mapped while preserving project-specific configurations. It includes a self-healing loop to resolve manual migration requirements and ensures that generated Codex artifacts are validated against target standards.

What it does

  • Converts Claude Code instruction files into Codex-compatible formats.
  • Migrates slash commands and skills into the Codex skill structure.
  • Updates MCP server configurations for use in Codex.
  • Generates a migration report detailing added items and manual review requirements.
  • Validates generated Codex artifacts to ensure configuration integrity.
  • Provides a dry-run mode to preview changes before applying them.

When to use it

  • When transitioning a development environment from Claude Code to Codex.
  • When you need to port custom skills or agents to a new Codex project.
  • When updating global or project-level configurations to align with Codex standards.
  • When you need to verify the readiness of existing Claude artifacts for Codex.

Advantages

  • Automates complex migration steps, reducing manual effort.
  • Provides clear reporting on what was successfully migrated and what requires manual attention.
  • Includes a self-healing loop to iteratively resolve migration issues.
  • Ensures project-specific configurations are preserved during the transition.
  • Validates targets post-migration to prevent configuration errors.

Usage examples

“Run a scan to see what Claude artifacts are available for migration.”
The tool lists active and inactive source surfaces from the specified Claude directory.
“Perform a dry-run migration of my project settings to Codex.”
The tool generates a plan and reports potential changes without modifying any files.
“Execute the full migration of my global Claude configuration to Codex.”
The tool migrates instructions, skills, and MCP configs, creating a report of the results.
“Validate the Codex target after my migration.”
The tool checks the generated Codex files for errors and reports any remaining manual fixes.

Install

  1. Download the skill as a ZIP:

    https://mp-staging.autogineer.com/agent-skills/openai/migrate-to-codex/download
  2. Unzip it into the cross-client skills folder of your project:

    mkdir -p .agents/skills && unzip migrate-to-codex.zip -d .agents/skills

Documentation by OpenAI

Migrate to Codex

Autonomy

Keep going until the selected migration is completely done: run the migrator, inspect the report, fix migrated Codex instructions/skills/agents/MCP config, and re-run checks without stopping to ask for confirmation of the next step. If the user has selected a target, do not ask before creating, editing, replacing, or deleting generated Codex artifacts in that target (AGENTS.md, .codex/, .agents/, or ~/.codex/). Preserve unrelated existing Codex config entries in .codex/config.toml or ~/.codex/config.toml, such as notify, projects, marketplaces, or unrelated MCP servers; do not ask about them unless they fail validation or directly conflict with the migration. Do not edit source Claude Code files (.claude/, ~/.claude/, .mcp.json, or .claude.json), unrelated project code, secrets, or another repository.

Migration Order

Run the migration in this order for each selected global or project source:

  1. Start by using Codex's built-in TODO/task list tool. Do not create MIGRATION_TODOS.md or any TODO file unless the user explicitly asks. The TODO list input has a plan array whose items each have step and status; use statuses pending, in_progress, and completed. Make the TODOs specific to the selected artifacts. Before finishing, update the TODO list so every finished step is marked completed and no step remains in_progress. Use literal source → Codex target labels, for example:

    • Inspect .claude/commands → Codex skills/prompts

    • Inspect .claude/agents → .codex/agents

    • Inspect .mcp.json → .codex/config.toml MCP servers

    • Inspect .claude/settings.json hooks → .codex/hooks.json

    • Migrate safe selected artifacts → Codex files

    • Validate generated .codex/config.toml

    • Validate generated .codex/agents

    • Report migrated artifacts and manual-review items

  2. Read references/differences.md (and refresh Codex docs if its Docs last checked date is old).

  3. Scan and inspect before writing:

    • --scan-only lists active and inactive source surfaces.

    • --plan prints staged Codex artifact paths and report rows.

    • --doctor summarizes readiness, manual-review work, and validation risks.

  4. Convert surfaces in the same order the CLI uses:

    • instructions: CLAUDE.md / AGENTS.md to AGENTS.md

    • plugins: report Claude plugin trees and marketplaces as manual migration work

    • hooks: rewrite supported Claude hooks into .codex/hooks.json and enable [features].codex_hooks = true

    • skills and commands: write Codex skills under .agents/skills/

    • config: write .codex/config.toml from Claude model/sandbox settings and MCP servers, including personality = "friendly" when config is generated

    • subagents: write Codex custom agents under .codex/agents/

  5. Dry-run, then write the selected target. Use --replace only when orphan generated skills or agents should be deleted.

  6. Inspect the terminal output and .codex/migrate-to-codex-report.txt after real runs.

  7. Review generated artifacts in this order: AGENTS.md, .agents/skills/, .codex/config.toml, .codex/hooks.json, .codex/agents/, then report-only plugin items.

  8. Run --validate-target against each target after edits.

  9. Re-run checks and --dry-run after edits.

  10. Return the final migration report as one markdown table per scope that has rows. The tables cover only the non-native follow-up migration work you performed, such as skills created from slash commands, subagents, MCP servers, hooks, unsupported/local plugin notes, and manual-review caveats. Include programmatic native import rows for config, instructions, skills, or supported plugins only if you personally migrated them in this follow-up run.

    If only one scope has rows, render only the table with no heading. If multiple scopes have rows, render one heading before each table. Use **User Config** for user-scope rows. For project-scope rows, use the actual project folder name as the heading, for example **northstar-support-portal**; do not use Current Project as the heading. Do not add prose before or after the table output.

    Use exactly these columns:

    northstar-support-portal

    StatusItemNotes
    AddedSlash command pr-reviewConverted into a Codex skill
    AddedSubagent release-leadAdded as a Codex subagent
    Check before usingHook PreToolUseConverted, but some Claude hook behavior differs in Codex
    Not AddedHook NotificationCodex does not have an equivalent notification hook
    Not AddedPlugin team-macrosPlugin needs manual setup

    Status must be Added, Check before using, or Not Added. Use Added when a Codex-facing artifact was created or changed and needs no special review. Use Check before using when a Codex-facing artifact was created or changed but the migration changed semantics, inferred behavior, preserved tool rules as guidance, or dropped unsupported behavior. Use Not Added when a source artifact was detected but no Codex-facing artifact was created. Item combines the artifact type and concrete item name in one cell. Artifact type must be singular: Skill, Slash command, Subagent, MCP, Hook, or Plugin. Wrap the artifact type in inline code; write the item name as plain text after it. Notes is always required; never leave it empty. Keep notes short, plain, and literal. Avoid internal implementation terms such as runtime expansion. Prefer phrases like Converted into a Codex skill, Added as a Codex subagent, Added to Codex config, Converted into a Codex hook, Converted, but some Claude hook behavior differs in Codex, Codex does not have an equivalent notification hook, Plugin needs manual setup, or Plugin marketplace needs manual setup.

Self-Healing Loop

Keep looping until the selected migration is complete:

  1. Run --plan or --doctor.

  2. Run the migration with --dry-run.

  3. Run the migration for real.

  4. Fix every generated ## MANUAL MIGRATION REQUIRED block and every manual_fix_required or skipped report row that can be resolved inside Codex artifacts.

  5. Run --validate-target.

  6. Re-run the migrator and validator until the report and validator have no actionable generated-artifact fixes left.

Do not edit source Claude Code files, unrelated project code, secrets, or another repository during this loop. If a report row requires source-provider changes or product judgment, leave the generated Codex artifact with clear manual guidance instead of changing the source.

Commands

Choose the migrator command.

MIGRATE_TO_CODEX='python3 .codex/skills/migrate-to-codex/scripts/migrate-to-codex.py'

Inspect the migration before writing.

$MIGRATE_TO_CODEX --source ~/.claude/ --scan-only
$MIGRATE_TO_CODEX --source ~/.claude/ --target ~/.codex/ --plan
$MIGRATE_TO_CODEX --source ~/.claude/ --target ~/.codex/ --doctor

Dry-run, then run without --dry-run, for global and project.

$MIGRATE_TO_CODEX --source ~/.claude/ --target ~/.codex/ --dry-run
$MIGRATE_TO_CODEX --source ~/.claude/ --target ~/.codex/
$MIGRATE_TO_CODEX --source ./.claude/ --target ./.codex/ --dry-run
$MIGRATE_TO_CODEX --source ./.claude/ --target ./.codex/

Run the post-migration validator against each target after edits.

$MIGRATE_TO_CODEX --validate-target ~/.codex/
$MIGRATE_TO_CODEX --validate-target ./.codex/

Run $MIGRATE_TO_CODEX --help for flags (--scan-only, --plan, --doctor, --validate-target, defaults, and so on). Deep tables and more links are in references/differences.md.

Files

references/differences.md

Migration Differences

Summary

This reference covers Claude Code to Codex migration only. It lists migration differences, partial mappings, and unsupported Claude Code behavior. Direct 1:1 mappings are intentionally omitted. When the converter preserves Claude-only semantics as prompt guidance, it also emits a manual_fix_required report row and writes a ## MANUAL MIGRATION REQUIRED block into the generated file.

Docs last checked: 2026-04-20. If today's date is later, re-open the official Codex docs below and the Claude Code docs map before trusting these mappings.

Instructions

SourceCodexMigration behaviorCaveat
.claude/CLAUDE.md, CLAUDE.md, or claude.mdAGENTS.md symlinkLinked automatically when content looks provider-neutralThis keeps one shared instruction body instead of duplicating docs.
Root AGENTS.mdRoot AGENTS.mdReported as activeThe converter does not overwrite or symlink the target file to itself.
Instruction content with /hooks, .claude/agents/, settings paths, subagent language, or permission-mode assumptionsGenerated AGENTS.md copyManual rewrite passThe converter intentionally breaks the symlink when obvious Claude-only semantics need a Codex-specific edit.

Commands

SourceCodexMigration behaviorCaveat
.claude/commands/*.md.agents/skills/source-command-<name>/SKILL.mdConverted to one-file Codex skillsSlash-command invocation, argument-hint, allowed-tools, $ARGUMENTS, shell-output interpolation, and file-reference expansion are preserved as manual-review text.
Command files with runtime expansionOne-file Codex skills plus manual_fix_required rowsPreserved as prompt textArgument placeholders, shell-output interpolation, automatic file expansion, model/agent routing, and executable hook behavior have different runtime behavior and must be checked manually.

Skills

SourceCodexMigration behaviorCaveat
.claude/skills/<name>/SKILL.md.agents/skills/<name>/SKILL.mdConverted; selected support directories are copiedSkill-local scripts/, references/, and assets/ are copied when they are real files under the source skill root.
.claude/skills/<name>.md.agents/skills/<name>/SKILL.mdConverted as a single-file skillNo sibling support directories are copied for this legacy shape.
allowed-toolsNo strict skill allowlistPreserved as prompt guidance in SKILL.mdagents/openai.yaml can declare tool dependencies, but that is not a permission boundary.
user-invocablepolicy.allow_implicit_invocationManual review onlySimilar intent, not equivalent semantics.
model / effortNo skill-level model pinUnsupportedCodex model selection is session/agent scoped in this converter.
disable-model-invocationNo direct equivalentUnsupportedRequires a manual rewrite if the source skill depends on this behavior.
argument-hint / context / agent / hooks / paths / shellNo direct equivalentUnsupportedKeep only if the behavior can be rewritten into prompt guidance or config.

MCP and config

SourceCodexMigration behaviorCaveat
.mcp.json or .claude.json mcpServers.codex/config.toml [mcp_servers.<name>]ConvertedProject .mcp.json and global .claude.json use the same source shape for this migrator. Codex supports additional MCP server fields such as cwd, enabled_tools, disabled_tools, and timeout settings, but this converter only writes fields that map clearly from Claude source config.
Claude Code model/sandbox settings or MCP configpersonality = "friendly"Written when the migrator generates .codex/config.tomlCodex supports none, friendly, and pragmatic; Claude Code migrations default to friendly to preserve a warm assistant style.
type: sseNo SSE supportUnsupportedCodex supports stdio and streamable HTTP in current docs.
headers.Authorization: Bearer ${VAR}bearer_token_env_varDirect auth rewriteOnly the bearer-token shape is rewritten this way; ${VAR:-default} fallbacks are not preserved.
headers with ${VAR}env_http_headersPartial mappingStatic headers map to http_headers; ${VAR:-default} fallbacks are not preserved.
env with ${VAR}env_varsPartial mappingLiteral values stay in env; self-references become env_vars, and ${VAR:-default} fallbacks are not preserved.
oauth.callbackPortmcp_oauth_callback_portManual review onlyoauth.clientId, oauth.authServerMetadataUrl, and headersHelper are unsupported.
enabledMcpjsonServers / disabledMcpjsonServersPer-server enabledPartial mappingenableAllProjectMcpServers has no direct equivalent in this converter.
allowedMcpServers / deniedMcpServersrequirements.tomlManual policy mappingNot written by this converter.
.claude/settings.local.jsonNo local-only Codex equivalentUnsupportedCodex project config is tied to trusted project behavior.

Subagents

SourceCodexMigration behaviorCaveat
.claude/agents/*.md.codex/agents/*.tomlConvertedMissing name or description is inferred and reported for review.
tools / disallowedToolsNo source-style fine-grained agent permissionsPreserved as prompt guidance in developer_instructionsUse sandbox_mode, [permissions], MCP tool filters, or app tool filters manually when intent is clear.
skillsNo spawn-time preload equivalentPreserved as prompt guidance in developer_instructionsskills.config is enable/disable config, not preload behavior.
mcpServersCodex custom-agent mcp_servers or shared Codex MCP configManual review onlyCodex custom-agent files can include MCP config, but this converter does not automatically map Claude subagent mcpServers. Use shared Codex MCP config or manually add agent-local mcp_servers when the source intent is clear.
permissionModesandbox_modePartial mappingOnly acceptEdits and readOnly are mapped; default, dontAsk, bypassPermissions, and plan are preserved as manual-review prompt guidance.
model + effortmodel + model_reasoning_effortPartial mapping by model familySonnet-family effort is biased one tier higher for coding-agent behavior; source max maps to Codex xhigh.
hooks / memory / background / isolation / maxTurnsNo direct equivalentUnsupportedForeground/background and resume behavior do not map cleanly to Codex custom-agent files.
initialPromptNo direct equivalentUnsupportedOnly applies when the agent runs as the main Claude session agent.
Auto-delegation by descriptionAutomatic or explicit Codex sub-agent spawningBehavior changeNot a 1:1 match; verify generated agent descriptions manually.
Independent agent permissionsParent sandbox inheritance + runtime overridesBehavior changeCodex custom-agent files set defaults, not hard isolation from the parent turn.

Plugin Marketplaces

SourceCodexMigration behaviorCaveat
.claude/plugins/Codex plugins / skills / MCP servers / appsReported as manual_fix_required onlyCodex plugins can bundle skills, MCP servers, and apps, but the migrator does not copy plugin trees. Migrate the plugin, bundled skills, commands, agents, hooks, and MCP config by hand.
.claude/plugin-marketplaces.jsonCodex plugin install or local plugin pathReported as manual_fix_required onlyMarketplace entries can point to local or remote plugin sources; the migrator does not fetch or install them. Codex marketplace metadata lives under .agents/plugins/marketplace.json or ~/.agents/plugins/marketplace.json.
.claude-plugin/marketplace.jsonCodex plugin install or local plugin pathReported as manual_fix_required onlyTreat it as marketplace source material. Do not copy it into Codex as a legacy marketplace; adapt it to the Codex plugin marketplace layout if you keep it local.
metadata.pluginRootNo direct equivalentUnsupportedShorthand plugin sources that depend on metadata.pluginRoot need manual layout.
Marketplace or plugin.json custom skills / agents pathsCodex plugin manifest and bundled skill pathsManual review onlyCodex plugins can declare bundled skills, MCP servers, and apps. Custom Claude plugin paths still need manual layout review; no automated scan.
Plugin commands/.agents/skills/<name>/SKILL.mdManualTreat like any other command migration if you copy files by hand.
strict, hooks, mcpServers, lspServers, outputStylesNo direct equivalentUnsupportedNo automatic plugin config import.

Hooks

SourceCodexMigration behaviorCaveat
hooks in ~/.claude/settings.json, .claude/settings.json, or .claude/settings.local.json.codex/hooks.json + [features].codex_hooks = truePartial conversionReview behavior before relying on migrated hooks; Claude and Codex hook runtimes are not 1:1.
NotificationnotifyManual rewrite onlynotify is a turn-complete notification command, not a general lifecycle hook or approval-prompt hook.
PreToolUsePreToolUse in .codex/hooks.jsonPartial conversionCodex currently runs PreToolUse for shell commands only and blocks only permissionDecision: "deny", legacy decision: "block", or exit code 2.
PostToolUsePostToolUse in .codex/hooks.jsonPartial conversionCodex currently runs PostToolUse for shell commands only; decision: "block" becomes model feedback, and continue: false stops execution. Formatting or fixups that Claude tied to Edit/Write should move to a Stop hook, because only Bash is matched for PostToolUse.
UserPromptSubmitUserPromptSubmit in .codex/hooks.jsonPartial conversionCodex can inject context or block a prompt, but it ignores matcher for this event and does not support source if filters.
SessionStartSessionStart in .codex/hooks.jsonPartial conversionCodex matches startup and resume; Claude may also expose other session flows.
StopStop in .codex/hooks.jsonPartial conversionCodex ignores matcher for Stop, can request a continuation prompt, and does not expose every source subagent/teammate stop lifecycle.
PermissionRequest / SubagentStart / SubagentStop / TaskCreated / TaskCompleted / StopFailure / PreCompact / PostCompact / SessionEndNo direct equivalentUnsupportedKeep as manual follow-up items; Codex does not expose matching lifecycle coverage today.
type: "command"type: "command"Partial conversioncommand, timeout / timeoutSec, and statusMessage map. Empty commands are skipped by Codex.
type: "prompt" / type: "agent" / type: "http" / async: trueNo direct equivalentUnsupportedCodex parses prompt / agent but skips them, and async hooks are skipped. HTTP hooks need a wrapper command.
Hook matcher + if filtersRegex matcher onlyPartial conversionCodex keeps regex matcher for PreToolUse, PostToolUse, and SessionStart only. Source if filters do not map.
Hooks in skills, agents, and pluginsNo direct equivalentUnsupportedCodex discovers hooks from config layers, not from skill or subagent manifests.

Planning and validation

CommandBehaviorCaveat
--planPrints staged migration output and generated artifact paths without writing filesStill depends on the selected source, target, and component flags.
--doctorPrints readiness, risk counts, and manual-review items without writing filesStatic guidance only; it does not prove the migrated setup works.
--validate-targetValidates an already migrated Codex targetChecks TOML parseability, skill frontmatter, custom-agent TOML fields, AGENTS.md size, and MCP command availability.

Minimal examples

Source skill metadata becomes prompt guidance:

allowed-tools:
  - Read
  - Bash
## MANUAL MIGRATION REQUIRED

Claude `allowed-tools` was preserved as prompt guidance, not a Codex permission boundary.

You're allowed to use these tools:

- Read
- Bash

Source subagent metadata becomes TOML plus prompt guidance:

skills:
  - release-notes
tools:
  - Read
disallowedTools:
  - Bash
sandbox_mode = "workspace-write"
developer_instructions = """
## Skills
- $release-notes

## Tools
You're allowed to use these tools:
- Read

Don't use these tools:
- Bash
"""

Sources

Requirements and compatibility

Written for: OpenAI Codex

Works with: Claude Code, Claude apps, OpenAI Codex, GitHub Copilot, Cursor, Gemini CLI, Kiro, OpenCode, Goose, Amp, Roo Code, JetBrains Junie, Factory Droid, Devin (ex-Windsurf), Mistral Vibe, OpenHands, Letta

“Works with” means the tool reads the open Agent Skills format; a skill written for one tool may use features others lack.

Needs: Needs an MCP server, Needs Python

Security

What our scanner noticed· Low risk

  • Scripts that your agent may run

This skill includes scripts that your agent may run. Read them before you use it.

A static scan is not a guarantee. Skills are third-party content: review a skill before you let an agent use it.

Attribution and licence

Published by
OpenAI
Publisher
OpenAI (Official)
Source
openai/skills · skills/.curated/migrate-to-codex · commit 49f948f
Licence
Apache License 2.0 (redistribution allowed)
  • ATTRIBUTION.md (written by AutoGineer)
Synced

AutoGineer is not affiliated with or endorsed by OpenAI.

Download ZIP

The ZIP holds the skill’s files, its licence files and an ATTRIBUTION.md. Download migrate-to-codex.zip