How to use Claude Code and Codex together in 2026: 10 ways compared

Disclosure: Enjoy is our product, and it runs both agents. We checked every way below against its maker’s own pages on November 1, 2026, and we say where another way is the better pick, including the free ones that don’t need Enjoy.

The ways to use Claude Code and Codex together, compared

Claude Code and Codex do the same job on the same files, and each has limits its own lab sets, so people pair them: one builds and the other reviews, or the second takes over when the first runs out. Our Claude Code vs Codex guide covers which to pay for. This one covers how to run both, and the ways differ mostly in how much of the work the second agent gets to see.

Way	Best for	Price, checked November 1, 2026	What the second agent gets	Runs on
Two agents side by side	Free, if you’re at home in a terminal	Nothing beyond your two plans	What you tell it	Mac, Windows, Linux
OpenAI’s Codex plugin	Codex reviewing or rescuing Claude Code’s work	Free; Codex usage comes from your ChatGPT plan or API key	The task you hand it, or the session with /codex:transfer	Wherever Claude Code and the Codex CLI run
Codex’s import	Moving your setup and chats to Codex	Free with Codex	Instructions, settings and chats from the last 30 days	ChatGPT desktop app, Codex CLI
Both in your editor	One window, with an agent in each panel	Nothing beyond your two plans; Zed’s Personal plan is $0	What you tell it	VS Code, Cursor and other forks, JetBrains IDEs, Zed
GitHub Copilot’s agents	Both on one plan, for tasks that end in a pull request	Copilot Pro+ $39 a month; Max $100	The issue or prompt	GitHub’s cloud
T3 Code	Switching one thread between them, free	Free	A selection of the thread, 16,000 tokens to start by default	Mac, Windows, Linux
Paseo	Agents that start each other, free	Free	The prompt one agent sends another	Mac, Windows, Linux
Orca	The same prompt to both, free	Free	The same prompt	Mac, Windows, Linux
Conductor	Reviewable workspaces, on a Mac	Free; Pro $50 a month; Teams $60 a user	The same branch and code	Mac
Enjoy	One conversation across both, in every project	A monthly subscription, free for 7 days, with your phone and other computers (pricing)	The conversation so far	Mac, Windows, Linux

Prices are per month as each maker lists them in the US, before tax. Apart from Copilot, every way needs Claude Code on a Claude plan or API key and Codex on a ChatGPT plan or API key, which those companies bill for separately. Claude’s pricing says “Claude Code is included in all paid plans,” and ChatGPT Plus lists “Codex on the web, in the CLI, in the IDE extension, and on iOS,” so the pair starts at $40 a month, or $37 with Claude Pro billed yearly.

If you’re on a Mac, Paradev, formerly Superset, has a handoff too: Continue with another agent “Starts a fresh session of the agent you pick, seeded with this terminal’s recent context,” the last 36,000 characters of it. Our Conductor alternatives guide covers Paradev and these apps in more depth.

Before you start

Four things apply to nearly every way below.

• One instruction file. Codex “reads AGENTS.md files before doing any work,” per its docs, and Claude Code “can read AGENTS.md as your project instructions,” but “only when you have no CLAUDE.md in your working directory or above it.” If you already have a CLAUDE.md, put @AGENTS.md at its top, as Anthropic’s docs show, and both agents get the same rules.
• Separate copies of the files. Two agents in one folder can overwrite each other’s edits. A Git worktree is a separate checkout on its own branch. claude --worktree feature-auth makes one for Claude Code, and in the ChatGPT app, worktrees “let Codex run multiple independent chats in the same project without interfering with each other.”
• Two sets of limits. Anthropic’s and OpenAI’s limits are separate, so one can run out while the other still works. Claude’s “are shared across Claude and Claude Code,” says its help center, and “ChatGPT Work and Codex share usage.” A pair doesn’t add up to one bigger limit: each agent spends its own plan.
• Skip the MCP bridge older guides describe. OpenAI’s command reference says “The codex mcp-server command and standalone codex-mcp-server binary have been removed,” and points to its app server. Claude Code’s claude mcp serve is still there, but its docs say it “only exposes Claude Code’s tools to your MCP client,” and don’t describe it as a way to hand Claude a task.

Two agents side by side

Best for: people at home in a terminal who want nothing extra.

Open Claude Code in one terminal and Codex in another, each in its own worktree. One writes, and you ask the other to read the diff. To hand a task across, ask the first to write down the goal, what’s done and what’s left in a file, and start the second by pointing at it. Each can also call the other from its command line, since claude -p and codex exec both run without their interfaces, so you can ask Claude Code to run codex exec "review the last commit".

Price: nothing beyond your two plans.

Watch for: neither agent sees the other’s conversation, so what you carry across is what you wrote down. Codex “stops adding files once the combined size reaches the limit defined by project_doc_max_bytes (32 KiB by default),” and Claude Code doesn’t read AGENTS.override.md, which Codex does.

OpenAI’s Codex plugin, inside Claude Code

Best for: having Codex check Claude Code’s work, or take on a stuck task, without leaving Claude Code.

OpenAI’s plugin lets you “Use Codex from inside Claude Code for code reviews or to delegate tasks to Codex.” It’s open source under Apache 2.0, with more than 34,000 stars. In Claude Code, run /plugin marketplace add openai/codex-plugin-cc, /plugin install codex@openai-codex, /reload-plugins and /codex:setup. Then:

• /codex:review is “a normal read-only Codex review” of your uncommitted changes or your branch.
• /codex:adversarial-review “questions the chosen implementation and design.”
• /codex:rescue “Hands a task to Codex through the codex:codex-rescue subagent,” to investigate a bug or try a fix.
• /codex:transfer “Creates a persistent Codex thread from the current Claude Code session and prints a codex resume <session-id> command.”

Price: the plugin is free. It asks for a “ChatGPT subscription (incl. Free) or OpenAI API key,” and “Usage will contribute to your Codex usage limits.” It works through your local Codex CLI, which OpenAI’s pricing lists from Plus, so plan on Plus at $20 a month.

Watch for: it goes one way, and Anthropic’s docs list no matching plugin for Codex. The optional review gate “can create a long-running Claude/Codex loop and may drain usage limits quickly.” Its latest release, v1.0.6, came out on July 8, 2026, and its README’s examples still name older models such as gpt-5.4-mini. It needs Node.js 18.18 or later, and for a transfer, “The source must be under ~/.claude/projects.”

Codex’s own import

Best for: moving your setup and recent chats from Claude Code to Codex, with no plugin.

“The desktop app can import from Claude Code, Claude Cowork, or Cursor. Codex CLI can import from Claude Code or Cursor,” says OpenAI’s guide. In the ChatGPT app, open Settings > Import; in the CLI, type /import. Instruction files become AGENTS.md, settings.json becomes config.toml, and skills, plugins, MCP settings, hooks, subagents and “Chats from the last 30 days” come across. The app can also “keep imported work in sync with automatic updates,” and “Importing doesn’t change or delete your existing agent setup.”

Price: free with Codex.

Watch for: it’s one-way, into Codex. “Codex CLI imports up to 50 chats from the last 30 days,” and /import “isn’t available during a running task, in a remote session, or while connected to a local app-server daemon.” OpenAI says to “Review imported setup before you rely on it,” including “Hooks whose behavior may differ after import.”

Both in your editor

Best for: keeping the window you know, with an agent in each panel.

Anthropic’s extension “provides a native graphical interface for Claude Code, integrated directly into your IDE,” and Codex’s guide says “VS Code and compatible editors use the Codex extension; Xcode and JetBrains IDEs provide their own integrations,” naming Cursor and Windsurf. Both install side by side in VS Code and Cursor. Zed lists Claude and Codex among its external agents, each started as a thread in its Agent Panel.

Price: nothing beyond your two plans, though Codex in an editor starts with Plus. “Zed does not charge for External Agents,” and its Personal plan is $0.

Watch for: the two panels don’t talk to each other, so a handoff is still you, and with Claude Code in VS Code, “Some features are only available in the CLI.” In Zed, “Billing, legal terms, retention, and data handling are between you and the agent provider.”

GitHub Copilot’s Claude and Codex agents

Best for: both agents on one GitHub plan, for tasks that end in a pull request.

“You can use third-party coding agents alongside Copilot to work asynchronously on your development tasks,” say GitHub’s docs. You assign an issue or give a prompt from a repository’s Agents tab, from GitHub Mobile or from VS Code, or mention an agent in a pull request. Anthropic’s agent “can be powered by your existing Copilot subscription,” and so can OpenAI’s, so neither needs a Claude or ChatGPT plan.

Price: GitHub’s plans page lists “Delegate tasks to third-party coding agents like Claude by Anthropic and OpenAI Codex (Preview)” from Pro+, at $39 a month, and Max is $100. “Coding agents consume GitHub Actions minutes and AI credits.”

Watch for: “Third-party coding agents are currently in public preview.” The model lists are shorter than the labs’ own plans: the docs list Claude Sonnet 5, Opus 5 and Fable 5.1, and GPT-5.3-Codex and GPT-5.4 nano for Codex, but not Opus 5.5 or OpenAI’s GPT-6 models. The work happens on GitHub, not on your computer.

T3 Code

Best for: switching one thread between them, free and open source.

T3 Code says “Orchestrate Claude Code, Codex, Antigravity, OpenCode, Cursor, and Grok from one surface. Bring your own subscription,” and “Switch models mid-thread.” Its docs say it “transfers conversation context when you switch providers.” It’s MIT-licensed, for Mac, Windows and Linux, with phone apps, and “Every agent thread writes to its own branch.”

Price: free. Its FAQ says “There’s no way to pay us to use T3 Code.”

Watch for: “A handoff is a budgeted selection, not an agent-written summary,” with an initial allowance of 16,000 tokens by default, and the agent can fetch the rest with T3 Code’s thread-reading tool. “A handoff does not copy the outgoing provider’s reasoning, tool-call state, or attachments.” Its README says “We are very very early in this project. Expect bugs.”

Paseo

Best for: agents that start and talk to each other across labs, free.

Paseo’s README says “Agents in Paseo can create worktrees, launch other agents, and talk to them, across providers.” Its orchestration docs add that they can “split research, implementation, and review between agents, including agents from different providers,” so Claude Code can start Codex to review its work. It runs on Mac, Windows and Linux, with phone apps.

Price: “Paseo is free and open source,” under Apache 2.0. “You need agent providers installed with your own credentials.”

Watch for: “Paseo runs a local server called the daemon that manages your coding agents,” so the machine running it has to stay on. The agent doing the delegating decides what to pass along, in the prompt it sends.

Orca

Best for: giving both agents the same task and keeping the better result, free.

Orca says “Run Claude Code, Codex, and any other coding agent in parallel, each in its own worktree,” and “Fan one prompt across 5 agents, compare, merge the winner.” It’s “Free and open source for macOS, Windows, and Linux.”

Price: free.

Watch for: by default, “Orca pre-fills each supported CLI’s permission-bypass flag for new launches,” --dangerously-skip-permissions for Claude and --dangerously-bypass-approvals-and-sandbox for Codex, reasoning that “worktrees are disposable,” so choose Manual under Settings → Agents → Agent Permissions for terminal agents, and check Settings → Chat → Behavior for its native chats, if you want to approve commands. The same prompt to both also spends from both plans.

Conductor

Best for: Mac users who want each agent’s work in a workspace they can review and merge.

“Claude Code, Codex, and OpenCode come bundled in Conductor,” says its harnesses page. “Each chat tab uses one harness and one selected model,” and “Multiple chats in one workspace share the same branch and current code state,” so a Claude Code chat and a Codex chat can work on one branch. Its homepage now leads with “Run a team of coding agents in the cloud.”

Price: its pricing has a free plan for “Local workspaces on your Mac.” Pro, at $50 a month, adds “Conductor Cloud and a bunch of cloud workspace hours,” and Teams is $60 a user a month.

Watch for: “Conductor is not available for Windows or Linux yet,” says its installation page. Its docs describe chats sharing a branch, not handing a conversation from one agent to another, though its homepage shows an agent using the Conductor API to start a session with another model. And locally, “When your machine shuts off or you close the app, sessions terminate.”

Enjoy

Best for: one conversation that moves between them, across all your projects, and carries on when one runs out.

Enjoy is a desktop app for Mac, Windows and Linux that runs Claude Code and Codex, along with Grok Build, Antigravity, Muse Code and local models through OpenCode, on the plans you already have. What it adds for these two:

• One conversation, either agent. Switch agents as often as you like: “The new agent gets the conversation so far, including replies from other agents, before it works on your message.” Plan with Claude Code, build with Codex, then have Claude Code review it, and “When you switch back to an agent, it continues where it left off.”
• Work in progress, too. While an agent is working, select Stop and move to: “The current agent stops, and the new one picks the reply up again with the conversation so far.”
• Limits handled. When one runs out, Enjoy switches to another of your accounts for that agent, or waits for the reset and carries on. Select Stop waiting to hand the conversation to the other agent instead.
• One set of instructions. Project instructions “apply to every conversation in the project, whichever agent you use,” and your CLAUDE.md and AGENTS.md files keep working.
• Your history, brought in. Enjoy imports a project’s past Claude Code and Codex conversations, so you don’t start cold.

Price: Enjoy is one monthly subscription, and the first seven days are free, with a card to start. It includes your phone and any browser and your other computers, and each teammate you invite needs a subscription of their own. See pricing.

Watch for: Enjoy doesn’t add Opus to a ChatGPT plan or GPT to a Claude one: each agent runs on its own lab’s plan, and Enjoy runs the Codex CLI, which OpenAI includes from Plus. A switch passes the messages, not the first agent’s reasoning or tool output: the newest ones up to about 200,000 bytes, with a note telling the agent how to read the rest. It doesn’t stop two conversations from editing the same files, so overlapping work needs a worktree. Enjoy runs Claude Code headless, so if the reports in issue #97074, still open, are right, Claude Code in Enjoy uses the five-hour window faster than in a terminal. Waiting for a reset needs “Your computer must be on, with Enjoy open, when the usage resets,” and scheduled workflows also run only while it’s awake with Enjoy open. Enjoy isn’t open source, and it’s a monthly subscription on top of your AI plans once the free week ends. If switching is all you want, T3 Code does it free.

When one agent is enough

If you rarely run out, a second plan buys little, and you may not need a second lab for a second look. Anthropic’s best practices count “Getting a reviewer that hasn’t seen how the code was written” among the things a subagent is good for, on the plan you already have, though “You get parallelism or a fresh pair of eyes out of it, but not a smarter answer.” When a limit is the problem, our guides for Claude Code and Codex cover waiting, credits and the rest.

Add the second agent when one plan’s limits stop you most weeks, or when another lab’s review catches enough to be worth $20 a month. Start with the free ways: one AGENTS.md, a worktree each, and OpenAI’s plugin. Reach for an app when you’re tired of carrying the context across yourself.

How we compared

We read OpenAI’s plugin README and its Codex docs on importing, instructions, worktrees, commands, editors and pricing; Anthropic’s Claude Code docs, help center and pricing; GitHub’s docs and plans; Zed’s docs and pricing; and T3 Code’s, Paseo’s, Orca’s, Conductor’s and Paradev’s own pages and repositories on November 1, 2026, and every one is linked below. Licenses, star counts and release dates come from GitHub that day. We quote each maker’s own words for its limits and leave out what it hasn’t documented; the figure for what Enjoy passes on a switch comes from its own code. We didn’t run the plugin or the other apps ourselves. These tools change weekly, so we recheck this guide against its sources at least once a month, and the date at the foot of the page says when we last did.