Skip to content
Menu

Guides

How OpenCode's limits work, when it runs on your own provider plans or keys

Checked on 4 November 2026 against the sources below.

OpenCode has no usage limits of its own. It is an open-source coding agent that sends every request to whichever model provider you connect, so the limit you hit is that provider's limit for the way you connected. There are four ways today. With a provider's API key, you pay for each token and the only limits are the provider's rate limits, with no 5-hour or weekly window. With a ChatGPT Plus or Pro sign-in, which OpenCode supports, requests count against your ChatGPT plan's own limits, up to a weekly cap per app that you set in ChatGPT. With a Claude Pro or Max subscription, there is no supported way: OpenCode's docs say plugins exist but "Anthropic explicitly prohibits this", and Anthropic's own page says the preferred way for other tools is an API key. With OpenCode's own paid options, you either pay as you go with budgets, or subscribe to OpenCode Go, which has 5-hour, weekly and monthly ceilings at 20, 50 and 100 per cent of a monthly dollar allowance. To see your usage, run opencode stats, or use a local tool such as ccusage that reads OpenCode's logs.

Each way is set out below with the vendor pages it was checked against on 4 November 2026.

What OpenCode is, and why it has no limits of its own

OpenCode describes itself this way: "OpenCode is an open source AI coding agent. It's available as a terminal-based interface, desktop app, or IDE extension." You connect a model provider with the /connect command and either enter a key or, for some providers, sign in through the browser. OpenCode is the client; the model and the bill belong to the provider. That is why the question "how do OpenCode's limits work" has four answers, depending on what you connected.

Way 1: a provider's API key

For every provider, OpenCode's providers page offers the option to "Manually enter API Key". You then pay the provider's price per token: for example, Anthropic charges $4 per million input tokens and $20 per million output tokens for Claude Opus 5.5 (Anthropic's API pricing), and OpenAI says that with a key you "pay for Codex usage based on API pricing" (OpenAI's Codex pricing). There is no 5-hour or weekly window on an API key. The limits are the provider's rate limits and any spend limit you set in the provider's console, which is also where the bill shows up.

Way 2: a ChatGPT Plus or Pro sign-in

OpenCode's providers page lists a ChatGPT option for OpenAI: after /connect, you can "select the ChatGPT Plus/Pro option and it'll open your browser and ask you to authenticate". That is OpenAI's Sign in with ChatGPT. OpenAI's page on it says "Eligible ChatGPT Plus and Pro subscribers can also choose to use their ChatGPT plan for AI requests", that "App usage counts toward your existing plan limits", and that "An app's limit is a cap, not a separate allowance or a reserved portion of your plan."

You set the cap in ChatGPT: "open Settings > Usage to see how connected apps use your plan", then "Under App limits, open the percentage menu next to the app. Set its weekly limit as a percentage of your overall weekly plan usage." So on this route the limits are your ChatGPT plan's limits. OpenAI's pricing page describes those as a weekly limit, plus a 5-hour limit on Plus, and says "Pro plans currently have no five-hour limit". OpenCode's share is capped at the percentage you chose, and heavy use of ChatGPT itself or of Codex can use up the plan before OpenCode reaches its cap.

OpenAI's help article "Using your ChatGPT plan in other apps and sites", read in a browser on 4 November 2026, adds three details. Using your plan in participating commercial apps needs a Go, Plus or Pro plan, while "Supported open-source tools remain available to all ChatGPT users". The per-app limit "is a cap, not a separate pool of usage", so an app can hit its own limit while your plan still has usage left. And after your plan runs out, an app can only use your ChatGPT credits if you have opted in and set that app's limit to 100 per cent; that opt-in is off by default.

OpenAI's developer page for open-source apps explains that an open-source or locally hosted app "can request permission to use the user's ChatGPT plan for eligible Responses API requests", and that paid or remotely hosted apps must fill in an interest form first. The page carries no date.

Way 3: a Claude Pro or Max subscription, which is not supported

OpenCode's providers page is blunt: "There are plugins that allow you to use your Claude Pro/Max models with OpenCode. Anthropic explicitly prohibits this." It adds that "Previous versions of OpenCode came bundled with these plugins but that is no longer the case as of 1.3.0." For Anthropic, it offers the API key route only.

Anthropic's own account article says: "The preferred way to access Anthropic services using third-party software, tools, or services (“third-party tools”), including open-source projects, is through API key authentication through Claude Console or a supported cloud provider." It adds that Anthropic "may at its discretion allow paid subscribers who have enabled usage credits to use certain third-party tools", and that tools which "misrepresent their identity to Anthropic's servers, attempt to route third-party traffic against subscription limits, or otherwise violate applicable terms or policies" are prohibited.

In plain terms, if you want Claude models in OpenCode today, use an Anthropic API key and pay per token. A Pro or Max subscription's 5-hour and weekly windows are for Claude Code and claude.ai, and both vendors describe a plugin that spends them through OpenCode as against Anthropic's rules.

Way 4: OpenCode's own paid options

OpenCode's pay-as-you-go service is described on its Zen page: "You are charged per request and can add credits to your account", and "A processing fee of 4.6% + $0.31 is added to each credit purchase". It lists models at per-token prices, for example Claude Opus 5.5 at $4.00 per million input tokens and $20.00 per million output tokens. Its budgets "cap monthly spend for the workspace and each member".

OpenCode Go is a subscription to open coding models: Go costs $10 a month, Go Plus $40 a month, and Go Teams $20 a member a month, with each member getting their own allowance. Usage is measured against a monthly dollar allowance for each model, with three ceilings: "5-hour — 20% of the monthly limit; weekly — 50%; and monthly — 100%." Each response reports how much the request counted in a field called opencode.allowance, and "Go Teams starts with Extra Usage on."

These are the only routes where OpenCode itself sets a limit, and the only ones where OpenCode's own console shows your spending.

How to track your OpenCode usage

The slash commands inside OpenCode's terminal interface include no usage or cost command. On the command line, though, opencode stats will "Show token usage and cost statistics for your OpenCode sessions", with options to choose the number of days and to break the figures down by model, tool or project.

Local tools also read OpenCode's logs. ccusage lists OpenCode as a source (ccusage opencode daily) beside Claude Code and Codex. Where the bill lives depends on the route: the provider's console for an API key, Settings > Usage in ChatGPT for a ChatGPT sign-in, and the OpenCode console for Zen or Go.

For a team, this means an engineer on OpenCode may be spending through four different bills, none of which knows about the others. OpenReason Intel, a dashboard for Claude Code and Codex usage, does not support OpenCode yet; support is coming.

What to do next

  1. Read next

  2. See what used up your limit

    Session Forensics is a free page that reads the Claude Code and Codex session files already on your computer, in your browser. It lists the limits you hit and the turns that used each one up. There is nothing to install and no sign-up, and your files are never uploaded. For Claude Code its 5-hour figure is an estimate, because Claude Code's files do not record the meter.

    Would you rather have your own agent do it? Paste this prompt into Claude Code or Codex.

    Open Session Forensics (opens in a new tab)

  3. Get an email when the limits change

    Get an email when Claude Code or Codex limits change. One email per change, at most one a week, from OpenReason. Unsubscribe from any email in one click.

  4. Track it over time

    OpenReason Intel keeps a record of your Claude Code and Codex 5-hour and weekly windows from every machine you use, with each turn's estimated share of the window and what it cost once the window was full. It is free for small teams: sign up with Google, no card, then run one command on Mac, Windows or Linux.

    It shows you what happened; it does not stop a session as you work, and it cannot change your vendor's limits.

    Sign up free with Google

Questions people ask

How do OpenCode's limits work, given it runs on the user's own provider plans?
OpenCode sets no limits itself. With an API key, the limits are the provider's rate limits and you pay per token. With a ChatGPT Plus or Pro sign-in, requests count against your ChatGPT plan, up to a weekly cap per app that you set in ChatGPT. Claude Pro and Max subscriptions are not a supported route. OpenCode Go has its own 5-hour, weekly and monthly ceilings at 20, 50 and 100 per cent of a monthly dollar allowance.
How do Codex's 5-hour and weekly limits work?
OpenAI's pricing page describes a weekly limit on paid plans and a 5-hour limit on Plus and Business Standard seats, and says Pro plans currently have no 5-hour limit. When OpenCode uses your ChatGPT plan, its requests draw on those same limits, capped at the share you set for the app. The guide on Codex's limits has the detail.
How do Claude Code's 5-hour and weekly limits work?
They are windows on Claude's subscription plans, sized as multiples of the Pro plan and shared across Claude Code and claude.ai. They do not apply to an API key, which is the route OpenCode supports for Anthropic. The guide on Claude Code's limits has the detail.
How do I check my Codex usage?
Run /status in the Codex CLI or open the usage dashboard at chatgpt.com/codex/settings/usage. If you connected OpenCode with a ChatGPT sign-in, Settings > Usage in ChatGPT also shows how much of your plan each connected app used.

Sources

Every figure on this page was read from these pages on 4 November 2026. Limits and prices change; if a vendor page now says something different, the vendor page is right and this one is out of date.