Blog

OpenCode on your ChatGPT plan: setup, limits and what to watch

Run /connect, pick OpenAI, sign in with ChatGPT Plus or Pro. How OpenCode uses your plan, where the limits sit, how to cap it, and why Claude plans differ.

By BringYourSubs5 min read

Yes, OpenCode runs on your ChatGPT Plus or Pro plan: run /connect, choose OpenAI, pick the ChatGPT option, and sign in through your browser. OpenAI lists OpenCode among its approved open-source Sign in with ChatGPT integrations, and the requests then count toward your plan's Codex usage instead of an API bill.

We checked OpenCode's docs and repository and OpenAI's pages on 9 October 2026. Here is the setup, how the limits work, and the open issues worth knowing about this week.

Is OpenCode allowed to use your ChatGPT plan?

Yes, on both sides. OpenAI's Sign in with ChatGPT page names four open-source integrations: OpenClaw, OpenCode, Pi and T3. OpenAI's developer docs describe plan usage as an optional permission an open-source app can request, and say it "does not grant access to their ChatGPT conversations or other account context."

OpenCode's provider docs are just as direct: "We recommend signing up for ChatGPT Plus or Pro." The same page lists ChatGPT Plus among the subscriptions you can use "with zero setup".

OpenCode itself is free and MIT-licensed. Its GitHub repository had about 212,000 stars on 9 October 2026, and the latest stable release was v1.18.35, published on 6 October.

How do you connect OpenCode to your ChatGPT plan?

The steps from OpenCode's docs:

  1. Install OpenCode from opencode.ai and start it in your project folder.
  2. Run /connect and select OpenAI.
  3. Choose the ChatGPT Plus/Pro option. Your browser opens; sign in to the ChatGPT account that has the plan and approve the request to use your plan.
  4. Back in OpenCode, run /models and pick a model. OpenCode's docs say "all the OpenAI models should be available" at that point.

Credentials you add with /connect are stored in ~/.local/share/opencode/auth.json, per the docs. Treat that file like a password.

The menu labels are changing. On 29 September, the day OpenAI opened plan usage to other apps, OpenCode merged a separate Sign in with ChatGPT method alongside its older Codex sign-in. A 5 October change renamed the older ones to ChatGPT browser (legacy) and ChatGPT device code (legacy) so they are not confused with Sign in with ChatGPT. Depending on your version you may see the docs' "ChatGPT Plus/Pro" label or the newer names. On a laptop, pick Sign in with ChatGPT where you have it; on a server or over SSH, the device-code option lets you sign in from another device.

If you have business data rules: the docs say that for ChatGPT OAuth, OpenCode "automatically applies a regional inference residency requirement when one is advertised by your workspace credentials."

How much of your plan does OpenCode use?

OpenCode has no allowance of its own. OpenAI's partner page is clear: "App usage counts toward your existing plan limits. Connecting an app does not add a new allowance." OpenCode, the Codex app, the Codex CLI and any other connected app all draw on the same Codex and ChatGPT Work usage.

What that allowance looks like, from OpenAI's Codex pricing page on 9 October 2026:

Plan Price Five-hour window Notes
Plus $20/month Estimated 15 to 160 local messages on GPT-6.1 Sol, 350 to 3,000 on GPT-6 Luna Weekly limits may also apply
Pro $100, $200 or $500 a month "Pro plans currently have no five-hour limit" Weekly limits still apply

OpenAI stresses that these are estimates, not fixed limits, and that "model choice, context, reasoning, tool use, retrieval, and caching all affect usage". Two of its tips apply directly to OpenCode: keep your AGENTS.md small, and "limit the number of MCP servers you use", since every server adds context to each message.

When you run out, Plus and Pro users can buy credits to keep going. OpenAI's usage settings also have a switch, "Allow other apps to use credits after reaching your usage limit", which decides whether connected apps like OpenCode can spend those credits.

How do you cap or disconnect OpenCode?

OpenAI's controls live in ChatGPT, not in OpenCode:

  • Cap it. In ChatGPT, open Settings, Usage. Under App limits, set the app's weekly limit as a percentage of your plan and save. OpenAI notes that this is "a cap, not a separate allowance or a reserved portion of your plan."
  • Cut it off. Signing out locally is not the whole job. Go to Settings, Security and login, Sign in with ChatGPT, select the app and choose Disconnect.

The per-app controls apply to apps you allowed to use your plan through Sign in with ChatGPT. If you connected with one of OpenCode's legacy Codex methods, check that the connection shows up there before relying on a cap.

What issues should you know about this week?

OpenCode moves fast, and the new ChatGPT route is days old. Open and recent items on GitHub, as of 9 October 2026:

  • Prompt caching on the new route. A user reported on 7 October that in OpenCode 2.0.24 the Sign in with ChatGPT route reused only 3.44% of input tokens from cache in one session, with 14 of 23 calls getting no cache hits (issue #53765). Less caching means more of your plan per message. A fix merged the same day sends the session headers the Codex route uses; its own description calls it an experiment that still needs a live check.
  • Models vanishing. Issue #52363, opened 30 September and still open, reports OpenAI models disappearing from the picker when OpenCode Zen or Go is connected too.
  • GPT-5.5 leaves on 14 October. OpenAI retires GPT-5.5 from ChatGPT and Codex plans that day. OpenCode's new method refreshes your account's model list every 15 minutes, per the PR, but if you pinned gpt-5.5 in your config, change it. See what changes for apps on your plan.

OpenAI also lists what plan-usage apps cannot do yet. Its preview limitations page says image generation, file search, Code Interpreter, native computer use and hosted MCP connectors are not supported on this route. OpenCode's own tools run locally, so day-to-day coding is not affected, but don't expect ChatGPT's hosted tools inside OpenCode.

Can you use your Claude subscription in OpenCode?

No. OpenCode's docs say: "There are plugins that allow you to use your Claude Pro/Max models with OpenCode. Anthropic explicitly prohibits this." OpenCode stopped bundling those plugins as of version 1.3.0. We don't recommend them.

Your options on a Claude plan are to use a Claude API key in OpenCode, billed per token, or to run Claude Code itself, or an app that drives it, such as Conductor, T3 Code or Cline. Our guide Can I use my Claude subscription in other apps? explains the rule.

What are the alternatives to OpenCode on a ChatGPT plan?

  • Kilo Code if you want the same engine in an editor. Its Open VSX changelog shows it adopting OpenCode releases, and it connects to your ChatGPT plan through Kilo's BYOK page.
  • OpenClaw and Hermes Agent for a personal assistant rather than a coding agent.
  • Conductor, whose site says it runs the first-party Codex, Claude Code, Cursor and OpenCode agents, to run several of them side by side.
  • Codex itself, OpenAI's own agent, on the same plan.

For the wider picture, see open-source agents you can run on your own subscription and every app on works with ChatGPT. If you came here from Cursor, read Can you use your ChatGPT subscription in Cursor?.

Questions people ask

Can OpenCode use my ChatGPT Plus or Pro subscription?
Yes. OpenCode's provider docs say to run /connect, select OpenAI and choose the ChatGPT Plus/Pro option, which opens your browser to sign in. OpenAI lists OpenCode as one of four open-source Sign in with ChatGPT integrations, so this is an approved route, not a workaround.
Does OpenCode on my ChatGPT plan cost extra?
No. OpenCode is free and MIT-licensed, and requests through your plan count toward your existing ChatGPT and Codex usage instead of an API bill. OpenAI says connecting an app does not add a new allowance, so OpenCode shares your plan limits with Codex and every other connected app.
How do I stop OpenCode from using up my plan?
In ChatGPT, open Settings, Usage and set a weekly limit for the app under App limits, as a percentage of your plan. To cut it off completely, go to Settings, Security and login, Sign in with ChatGPT, select the app and choose Disconnect.
How do I sign in to OpenCode with ChatGPT on a remote server?
Recent OpenCode builds keep a device-code sign-in for machines without a browser. OpenCode's pull request of 5 October 2026 renamed it ChatGPT device code (legacy), next to the newer Sign in with ChatGPT method. Run /connect, choose OpenAI and pick the device-code option.
Can I use my Claude subscription with OpenCode?
No. OpenCode's docs say Anthropic explicitly prohibits the plugins that add Claude Pro or Max to OpenCode, and that OpenCode stopped bundling them as of version 1.3.0. Use a Claude API key in OpenCode, or run Claude Code itself on your Claude plan.

Sources

  1. Providers (OpenAI section), OpenCode, 2026-10-09
  2. anomalyco/opencode (repository and releases), GitHub, 2026-10-09
  3. feat(core): add separate ChatGPT token-sharing OAuth method (PR #52147), GitHub (anomalyco/opencode), 2026-09-29
  4. fix(core): clarify legacy ChatGPT OAuth method labels (PR #53451), GitHub (anomalyco/opencode), 2026-10-05
  5. V2: ChatGPT Token Sharing reports 3.44% prompt-cache reuse in native CLI (issue #53765), GitHub (anomalyco/opencode), 2026-10-07
  6. fix(core): send Codex session headers with ChatGPT token sharing (PR #53767), GitHub (anomalyco/opencode), 2026-10-07
  7. OpenAI models disappear when OpenCode Zen or Go is connected (issue #52363), GitHub (anomalyco/opencode), 2026-09-30
  8. Sign in with ChatGPT (partner list, usage and app limits), OpenAI, 2026-10-09
  9. Sign in with ChatGPT: ChatGPT plan usage overview (open source), OpenAI, 2026-10-09
  10. Sign in with ChatGPT: Preview behavior and limitations, OpenAI, 2026-10-09
  11. Pricing (ChatGPT Work and Codex), OpenAI, 2026-10-09
  12. Kilo Code changelog on Open VSX, Open VSX, 2026-10-09