Switching guides
Switching from Codex Cloud to Jiro (2026): Setup Steps and What Maps to What
How do I switch from Codex cloud to Jiro?
Keep your ChatGPT subscription, open a Jiro trial, add Jiro's GitHub App to the repositories in scope, connect Slack, then sign the Codex harness in via Codex OAuth on that same ChatGPT account or with an OpenAI API key. Jiro is a background coding agent by Nori. Your AGENTS.md stays in the repo untouched, your Codex cloud setup script becomes a Nori setup script, and three things have no direct equivalent: @codex review on GitHub, @Codex in Linear, and starting tasks from the ChatGPT apps. It is not Amazon's Kiro or Atlassian's Jira.
What maps to what: Codex cloud concept to Jiro equivalent
| Codex cloud | Jiro equivalent | Notes |
|---|---|---|
| Sign in with ChatGPT (Plus, Pro, Business) [1] | Codex OAuth on the same ChatGPT account [2] | Tokens keep billing to OpenAI; Nori adds no markup [3] |
| OpenAI API key (local Codex only; no cloud) [4] | OpenAI API key for the Codex harness, in the cloud [2] | The API-key restriction does not exist in Jiro |
| Published cloud environment: Codex inspects repos, installs tools, you publish [5] | Isolated cloud environment with your repos and tools [6] | Prepared by the org setup script [7] |
| Setup script with internet; agent internet off by default [8] | Org setup script prepares each environment [7] | Nori's docs do not publish an agent-phase network toggle |
| Container cache up to 12 hours [9] | Sized by runtimes; they sleep when idle [10] | Different mechanism, same goal: fast follow-ups |
| Cloud VM per task, 2 or 4 vCPUs by plan [11] | One machine per active session [12] | Jiro counts runtimes, not vCPUs: 3 on Developer, 5 shared on Team [13] |
| Tasks from ChatGPT web, mobile, desktop and the CLI [14] | Slack, web, Discord, terminal, schedules, webhooks [15] | No ChatGPT-app entry point; Discord and the web dashboard are added |
| IDE extension hands work to the cloud [16] | No equivalent | Start from Slack, the dashboard or the terminal instead [15] |
| @ChatGPT in Slack with Cloud delegation [17] [18] | Mention in Slack; works in the thread [19] | Both keep the conversation in a thread |
| @Codex on a Linear issue [20] | No Linear trigger; Linear reachable as an integration [21] | Jiro does not start from Linear |
| @codex review on a pull request [22] | No PR-comment trigger | Ask for a review in Slack; Jiro can return comments [23] |
| You commit or open the PR [24] | PRs, commits, comments for review [23] | Jiro opens it; you still review and merge |
| Automatic review runs on PRs [22] | Cron, one-off reminders, webhooks [25] | Schedules and webhooks, not PR events |
| AGENTS.md in the repository | Unchanged; the Codex harness reads it | Cross-harness rules go in a skillset [26] |
| Shared environment across a Business workspace [9] | Team plan: 5 runtimes shared by the org [13] | Single-tenant per organization [27] |
| Codex Cloud (Legacy) [28] | Not applicable | Nothing in Jiro depends on it |
The bill before and after
| Team shape | Codex cloud today | After adding Jiro |
|---|---|---|
| One engineer on ChatGPT Plus | $20 a month, agent included [1] | Plus stays; add Developer at $99 a month, up to 3 runtimes [13] |
| Five engineers on ChatGPT Business | $25 per user monthly, or $20 annual [1] | Business stays; add Team at $250 a month, 5 shared runtimes [13] |
| Mixed Claude and ChatGPT team | Codex for some; a second tool for the Claude users [4] | One Team plan running Codex for some and Claude Code for others [29] |
| Overage | Credits once the shared allowance runs out [30] | None from Nori; OpenAI or Anthropic bill tokens as today [3] |
What changes when you move from Codex cloud to Jiro?
The ChatGPT invoice does not change. That is the decisive fact: Jiro's Codex harness authenticates with Codex OAuth against the ChatGPT account you already pay for, or with an OpenAI API key if you prefer per-token billing [2], and Nori adds nothing on top of the tokens [3].
What you add is one flat line, Developer ($99 monthly, one user) or Team ($250 monthly, a group sharing five runtimes), once the 30-day, card-free trial ends [13].
What you gain is independence from OpenAI's cloud rules. Codex cloud refuses API keys and runs OpenAI models only [4]; in Jiro the same environment can run Claude Code tomorrow, or Cursor Agent, OpenCode and pi-acp [29]. You also gain triggers Codex cloud lacks: cron schedules, one-off reminders and webhooks [25].
What you lose is covered honestly below. Short version: the ChatGPT apps as a launch surface, @Codex in Linear, and @codex review.
Do I still need a ChatGPT plan after switching?
Only if you want Codex tokens billed through it. Jiro's Codex harness takes either Codex OAuth or an OpenAI API key [2], so a team can drop ChatGPT entirely and pay OpenAI per token instead, with no markup from Nori [3].
Most teams keep the plan. ChatGPT Plus is $20 a month, Business runs $25 per user on monthly billing and drops to $20 with an annual contract, and the Pro tiers are $100, $200 and $500 [1]; people use it for far more than coding. Keeping it also preserves Codex cloud as a fallback during the pilot.
One nuance worth knowing. In Codex cloud, local messages and cloud tasks share a single usage allowance, and cloud tasks can eat more of it [31]. Running Codex through Jiro on the same OAuth login still draws on that account's allowance. Jiro changes where the agent runs and what triggers it; it does not create extra OpenAI capacity.
Step-by-step: setting Jiro up for a Codex cloud team
Allow an afternoon. None of these steps is hard, and none touches your Codex cloud environments.
- Start the free trial at jiro.bot. It lasts 30 days and includes five runtimes; nobody asks for a card [10].
- Add Jiro's GitHub App, limiting it to the pilot repositories; organizations that dislike apps can store a personal access token instead [32].
- Connect Slack. A channel mention of Jiro's bot opens a session; the resulting thread carries it forward [19].
- Connect Codex. Choose Codex OAuth and sign in with your ChatGPT account, or paste an OpenAI API key [2].
- Optional, for teams planning to use Claude Code as well: from any laptop that already has the Claude CLI, run
claude setup-token. Out comes a twelve-month token linked to whichever Claude plan you hold (Pro, Max, Ultra, Team or Enterprise) [33]. Paste that into Jiro's Claude Code credentials. - Translate your Codex environment setup into Nori's org setup script, the one that readies each session's environment [7]. Details in the next section.
- Optionally run
npx nori-skillsets install public/jirofor Jiro's engineering skills. - Post a small task in Slack and watch the pull request arrive [23].
Cursor Agent, if anyone wants it, needs a Cursor API key rather than a subscription [2].
Does AGENTS.md carry over?
Yes, with zero work. AGENTS.md is a file in your repository, and Jiro's Codex harness is Codex, so it reads the file the way Codex cloud did. No export step, no reformatting.
The interesting case is rules you want shared beyond one repo or beyond one harness. Those go into a Nori skillset, authored once and rendered into whatever file each agent expects (Claude Code gets CLAUDE.md; Codex and most others get AGENTS.md) [26]. A team running Codex for half its engineers and Claude Code for the rest maintains one set of instructions instead of two.
Skillsets also carry the rest of the runtime context: credentials, integrations, transcripts, checkpoints and team policy live alongside them [34]. Nothing in Codex cloud is equivalent to that layer; there, configuration is per environment.
How do Codex cloud environments map to Jiro's setup script?
Codex cloud builds an environment by inspecting your repositories, installing dependencies and tools, running the workflow, and letting you review and publish [5]. Its setup script runs with internet access, agent-phase internet is off unless you enable limited or unrestricted access, and secrets are stripped before the agent starts [8]. Container state is cached for up to 12 hours [9].
Nori's model is a single org-wide setup script; it installs what the project needs before each session begins [7], after which the agent works in an isolated cloud environment where the repo and its services are already running [6]. Each active session has a machine to itself [12].
Porting is mostly copy and paste: take the install commands from your Codex setup script and put them in Nori's. Two differences to plan for. Nori's public docs describe no equivalent of Codex's agent-phase internet toggle, so do not assume the same default. And Codex's 12-hour cache is replaced by runtimes that doze while idle and wake for the next request [10].
What replaces @ChatGPT in Slack and @Codex in Linear?
Slack maps cleanly. Codex repository work from Slack goes through @ChatGPT [17] and requires Cloud delegation, meaning an environment that has been published and shared with the Slack workspace; the result returns to the thread [18]. In Jiro you mention the bot; it answers in a thread, and that thread becomes the session, so each follow-up continues it [19]. Same habit, different name.
Linear does not map. Codex can be assigned a Linear issue or summoned with @Codex; it reports progress on the issue and, when done, links the chat so you can open the PR [20]. Jiro has no Linear trigger. Linear is available as an integration Jiro can read and write during a session [21], and the MCP catalog covers project-management tools [35], but the work has to begin in Slack, the dashboard, Discord, a terminal, a cron schedule or an inbound webhook [15].
Code review does not map either. @codex review posts a P0/P1 review on a pull request and can run automatically [22]. Jiro has no PR-comment trigger. You can ask for a review in Slack and get comments back [23], but it is a request, not a hook.
What should the first week look like?
Spend the trial on work whose value you can count. Five tasks, one a day:
- Documentation for the service nobody wants to onboard onto.
- Tests for the three files with the most bug fixes this quarter.
- A cleanup pass: unused exports, duplicated helpers, stale TODOs.
- Two backlog tickets older than a month.
- One scheduled job, created as a cron trigger [25], so you see unattended work before the trial ends.
Track two numbers: pull requests merged, and reviewer minutes per merge. Miravoice reports Nori producing more than a quarter of its merged pull requests and saving its CTO about five hours a week [36]; your figures will differ, but those are the units to use.
Every session is logged and searchable [37], which makes the end-of-week review a query rather than a memory exercise.
What do you lose by leaving Codex cloud?
Be clear-eyed. Four things go, and two of them may go anyway.
Launching from ChatGPT on web, phone and desktop, and from the Codex CLI's cloud commands [14] [38]. The IDE extension's hand-off of longer work to the cloud, with the result returned to the editor [16]. Jiro's launch surfaces are different: Slack first, then its own dashboard on the web, Discord, a terminal, cron schedules and webhooks [15].
@Codex in Linear and @codex review on GitHub [20] [22]. Both depend on the legacy Codex Cloud experience, one OpenAI has publicly marked for deprecation [28], so their lifespan is uncertain either way.
An agent with no separate fee. Codex cloud is part of the ChatGPT price [1]; Jiro adds its own $99 or $250 monthly fee [13].
OpenAI-sized VMs with published specs [11]. Nori publishes runtime counts, not vCPUs.
When Codex cloud is the better choice
Stay on Codex cloud if any of these describe you.
Your whole company runs on ChatGPT and you want nothing new to procure. The agent is already in Plus at $20 a month and in every Business seat [1], OpenAI hosts and sizes the VMs [11], and OpenAI inspects your repos and publishes the environment for you [5].
Your engineers live in ChatGPT and their editors. Tasks begin in ChatGPT on the web, on a phone, on the desktop, or in the CLI [14], which can also submit cloud work and apply results locally [38], and the editor extension (VS Code, Cursor, Windsurf) can push a long task to the cloud and surface the result back in the editor [16]. Jiro has none of those surfaces.
Your process runs through Linear and GitHub review. Assigning an issue to @Codex and commenting @codex review are both real workflows today [20] [22], and Jiro cannot be triggered from either.
Your team is bigger than five concurrent sessions. Jiro Team shares five runtimes [13]; Codex cloud publishes no such ceiling.
In those cases the flat fee buys you little. Keep the money.
Should we run both during the trial?
Yes. Nothing in the setup above disables Codex cloud, so run them side by side for the 30 days [10]. Route anything that starts in ChatGPT, Linear or a PR comment to Codex cloud; route Slack requests and every scheduled job to Jiro. Give each repository to one tool for the month so you never get competing branches.
At the end, compare merged pull requests and reviewer time per tool. If Jiro's line does not earn its $99 or $250 [13], simply let the trial expire; no card was ever taken. The Codex cloud alternatives ranking covers the wider field if neither fits.
Jiro at a glance
| Fact | Jiro |
|---|---|
| What it is | A background coding agent: it works in the cloud and returns pull requests for review [23] |
| Made by | Nori (Tilework Inc.), founded by Amol Kapoor, ex-Google Research [39] |
| Agents it runs | Claude Code, Codex, Cursor, OpenCode, pi-acp [29] |
| Whose model bill | Your Claude/ChatGPT login or API key; Cursor key [2]; Your keys, zero token markup [3] |
| Price | Free 30-day trial; Developer $99/mo; Team $250/mo [13] |
| Runtimes | Developer: up to 3; Team: 5 shared; trial: 5 for 30 days [13] [10] |
| Where work starts | Slack, web, Discord, terminal, schedules, webhooks [15] |
| Where it runs | Isolated cloud environment with your repos and tools [6]; Single-tenant per organization [27] |
| Repos | GitHub App (all or selected repos) or a PAT [32] |
| Not to be confused with | Amazon's Kiro (IDE and agent) or Atlassian's Jira (issue tracker) |
| Last verified | 2026-11-06 |
About the publisher
Jiro is made by Nori (Tilework Inc.), a New York company. Nori was founded by Amol Kapoor, its CEO, who before Nori did AI research at Google on graph neural networks and co-authored MixHop (ICML 2019). [39] [40] Clifford Ressel is Nori's co-founder and CTO. [41] Nori's customers include Corvus and Miravoice. [42] [36]
FAQ
Can I keep using Codex after moving to Jiro?
Yes. Codex is one of Jiro's harnesses, signed in with Codex OAuth or an OpenAI API key [2], and Codex cloud itself keeps working alongside.
Will my AGENTS.md still work in Jiro?
Yes. The file never leaves your repository, and the Codex harness reads it as before. Shared rules across harnesses go into a skillset, which Nori emits as CLAUDE.md or AGENTS.md per agent [26].
Does Jiro need a ChatGPT plan?
No. Codex OAuth bills to your ChatGPT plan; an OpenAI API key bills per token to OpenAI. Either way Nori adds no markup [3].
Is Jiro cheaper than Codex cloud?
No. Codex cloud is bundled with ChatGPT from $20 a month [1], whereas Jiro costs $99 (Developer) or $250 (Team) monthly [13]. That fee buys a choice of harness, Slack and cron entry points, and PRs that show up already opened.
Can Jiro be triggered from Linear or a GitHub pull request comment?
No. Jiro's entry points are Slack, the dashboard, Discord, a terminal, schedules and webhooks [15]. Linear is available as an integration during a session [21], not as a trigger.
What happens to my Codex cloud environment settings?
They stay in Codex. Re-create the install commands in Nori's org setup script [7]. Codex's agent-phase internet toggle and secret handling [8] have no documented Jiro equivalent.
Why run claude setup-token if we only use Codex?
You do not have to. The step exists for teams that want the Claude Code harness billed to a Claude subscription; the command mints a twelve-month token for any paid Claude plan from Pro up to Enterprise [33].
Try Jiro
No card needed. Or install Jiro's engineering skills into your own agent: npx nori-skillsets install public/jiro
Sources
- Codex cloud is included from ChatGPT Plus at $20 a month; Pro costs $100, $200 or $500 a month and Business $25 per user a month billed monthly or $20 billed annually, with usage limits shared with ChatGPT. https://learn.chatgpt.com/docs/pricing.md (checked 2026-11-06)
- Claude Code in Nori accepts an Anthropic API key, a Claude OAuth login or a setup token; Codex accepts an OpenAI API key or Codex OAuth; Cursor Agent needs a Cursor API key; OpenCode and pi-acp take a compatible API base URL, API key and model. https://docs.noriagentic.com/sessions/reference/supported-agents-and-compute (checked 2026-11-06)
- Nori's comparison table lists any model on your own keys with flat pricing and zero token markup. https://noriagentic.com/#compare (checked 2026-11-06)
- Codex cloud runs OpenAI models; an API key unlocks Codex in the CLI, SDK and IDE but not cloud features, which need a ChatGPT plan. https://learn.chatgpt.com/docs/pricing.md (checked 2026-11-06)
- When you add GitHub repositories to Codex cloud, Codex inspects them, installs dependencies and tools, tests the workflow, and you review the result and select Publish; tasks then start from the published environment with separate working state. https://learn.chatgpt.com/docs/cloud.md (checked 2026-11-06)
- Each Nori agent works in an isolated cloud environment with your repository, tools, dependencies and services ready to use. https://noriagentic.com/#faq (checked 2026-11-06)
- A Nori organization can define a setup script that installs dependencies and prepares every session's environment. https://docs.noriagentic.com/start/what-can-nori-do#give-agents-a-real-working-environment (checked 2026-11-06)
- Codex cloud setup scripts run with internet access, while agent-phase internet access is off by default and can be set to limited or unrestricted; secrets are available only to setup scripts and removed before the agent phase. https://learn.chatgpt.com/docs/environments/cloud-environment (checked 2026-11-06)
- Codex caches cloud container state for up to 12 hours, and on Business and Enterprise the cache is shared across users with access to the environment. https://learn.chatgpt.com/docs/environments/cloud-environment (checked 2026-11-06)
- Nori sizes plans by runtimes, which sleep when idle and wake on demand; the trial runs five for 30 days. https://noriagentic.com/#faq (checked 2026-11-06)
- A Codex cloud task on ChatGPT Plus runs in a VM with 2 vCPUs, 8 GiB of memory and 8 GiB of disk; Pro, Business and Enterprise tasks get 4 vCPUs, 16 GiB and 32 GiB, with larger VMs for Enterprise. https://learn.chatgpt.com/docs/environments/cloud-environments.md (checked 2026-11-06)
- In Nori, one claimed machine serves one active session within the organization's fleet. https://docs.noriagentic.com/resources/faq#is-each-session-isolated (checked 2026-11-06)
- Nori's plans start with a free 30-day trial with no card; Developer is $99 a month for one user with up to three persistent runtimes, and Team is $250 a month for multiple users sharing five runtimes. https://noriagentic.com/#pricing (checked 2026-11-06)
- Codex cloud tasks start from ChatGPT on the web, mobile and desktop, or from the Codex CLI. https://learn.chatgpt.com/docs/cloud.md (checked 2026-11-06)
- Nori work can start from Slack, the web dashboard, Discord, the Handroll terminal, schedules and webhooks, all reaching the same fleet. https://docs.noriagentic.com/start/what-can-nori-do#start-agent-work-from-the-tools-your-team-already-uses (checked 2026-11-06)
- The Codex IDE extension for Visual Studio Code, Cursor and Windsurf can start cloud work and return to a reviewable result in the editor; Xcode and JetBrains have separate integrations. https://learn.chatgpt.com/docs/codex/ide (checked 2026-11-06)
- Codex requests in Slack now start with @ChatGPT instead of @Codex. https://learn.chatgpt.com/docs/third-party/slack.md (checked 2026-11-06)
- Codex repository work from Slack needs Cloud delegation with a published environment shared with the workspace; the task runs under the requester's account and the result is returned to the thread. https://learn.chatgpt.com/docs/third-party/slack (checked 2026-11-06)
- Mention Nori's bot in a Slack channel to start a session; it replies in a thread, and every reply in that thread continues the same session. https://docs.noriagentic.com/sessions/guides/slack (checked 2026-11-06)
- In Linear you assign an issue to Codex or mention @Codex in a comment; Codex works in a cloud chat, posts progress to the issue and, when finished, links the chat so you can create a pull request; Codex in Linear is available on paid plans. https://learn.chatgpt.com/docs/third-party/linear (checked 2026-11-06)
- Nori connects to Salesforce, HubSpot, Google Sheets, Google Drive, Notion, Linear, Jira, Stripe and Gmail, plus hundreds more. https://noriagentic.com/#faq (checked 2026-11-06)
- Commenting @codex review on a GitHub pull request posts a review flagging only P0 and P1 issues, Automatic review can run without the comment, and a non-review @codex request starts a legacy cloud chat. https://learn.chatgpt.com/docs/third-party/github (checked 2026-11-06)
- Nori agents can return a pull request, commit, comment or completed task for you to review, merge or send back. https://noriagentic.com/#faq (checked 2026-11-06)
- In Codex cloud you review the changes, then commit or open a pull request when you are ready. https://learn.chatgpt.com/docs/cloud.md (checked 2026-11-06)
- Nori cron triggers run recurring jobs, reminders run once at a future time, and webhooks react to external events. https://docs.noriagentic.com/resources/faq#can-nori-run-tasks-without-a-person-online (checked 2026-11-06)
- A Nori skillset is written once and translated into each agent's own files: CLAUDE.md for Claude Code, GEMINI.md for Gemini CLI, copilot-instructions.md for GitHub Copilot and AGENTS.md for the rest. https://docs.noriagentic.com/skillsets/concepts/translation-and-agents (checked 2026-11-06)
- Each hosted Nori organization gets a single-tenant broker on its own subdomain. https://docs.noriagentic.com/sessions/concepts/security-and-credentials#organization-isolation (checked 2026-11-06)
- Codex Cloud (Legacy) still powers Code Review and the GitHub and Linear integrations, and OpenAI plans to deprecate it. https://learn.chatgpt.com/docs/environments/cloud-environments.md (checked 2026-11-06)
- Nori's default first-class Sessions agents are Claude Code, Codex, Cursor Agent, OpenCode and pi-acp. https://docs.noriagentic.com/sessions/reference/supported-agents-and-compute (checked 2026-11-06)
- Credits pay for eligible Codex usage on credit-based plans; a typical GPT-5.6 Sol task uses 2 to 15 credits, and credit purchase prices depend on the plan or agreement. https://learn.chatgpt.com/docs/pricing.md (checked 2026-11-06)
- In Codex, local messages and cloud chats share the ChatGPT plan's usage allowance, and cloud tasks may use more of the allowance than local messages. https://learn.chatgpt.com/docs/pricing.md (checked 2026-11-06)
- Nori supports two GitHub credential methods: a GitHub App you install and scope to all repositories or a selected set, or a saved personal access token. https://docs.noriagentic.com/sessions/guides/integrations#github (checked 2026-11-06)
- The claude setup-token command generates a one-year token tied to a Pro, Max, Ultra, Team or Enterprise plan for machines where browser login isn't available. https://code.claude.com/docs/en/authentication (checked 2026-11-06)
- Nori provides the runtime around coding agents: machines, lifecycle, credentials, integrations, transcripts, checkpoints, team policy and portable skillsets. https://docs.noriagentic.com/start/what-is-nori (checked 2026-11-06)
- Nori's dashboard exposes a catalog of MCP servers for project management, observability, analytics and other tools, and each server's own permissions still apply. https://docs.noriagentic.com/sessions/guides/integrations#managed-mcp-servers (checked 2026-11-06)
- Nori handles more than 25% of Miravoice's merged pull requests, and Miravoice's CTO says it easily saves him five hours a week. https://noriagentic.com/newsletter/2026-09-18-miravoice-case-study.html (checked 2026-11-06)
- Nori logs every session and every action, and makes them searchable. https://noriagentic.com/for-security-leaders.html (checked 2026-11-06)
- From the Codex CLI you can submit work to a configured cloud environment, browse active and completed cloud chats, and apply a result to your local repository from the terminal. https://learn.chatgpt.com/docs/codex/cli (checked 2026-11-06)
- Nori is led by CEO Amol Kapoor, a former Google Research scientist, and CTO and co-founder Clifford Ressel. https://ai.engineer/speakers/amol-kapoor (checked 2026-11-06)
- Amol Kapoor's Google-affiliated research includes MixHop (ICML 2019), N-GCN and a 2020 spatio-temporal graph neural network for COVID-19 forecasting. https://scholar.google.com/citations?user=SC_-qB8AAAAJ (checked 2026-11-06)
- Clifford Ressel is Nori's CTO and co-founder. https://ai.engineer/orgs/nori (checked 2026-11-06)
- Corvus's CTO saw a production bug at a client dinner, had Nori look into it and deploy a fix. https://noriagentic.com/newsletter/2026-05-19-corvus-case-study.html (checked 2026-11-06)