What is bring your own subscription (BYOS) for AI apps?

Bring your own subscription (BYOS) means an app runs its AI work on a plan you already pay for, such as Claude Pro or Max or ChatGPT Plus or Pro, instead of selling you its own AI credits or asking for a metered API key. The work comes out of that plan's normal usage limits, so you pay one flat AI bill and use it across several apps.

The term is new enough that Google still suggests it for enterprise software licensing (Autodesk, Red Hat). In AI it describes a specific, recent shift: in 2026 both OpenAI and Anthropic set out how other apps may use their consumer plans. This guide defines it, shows how it works on each provider, and lists what to check before you connect an app. BringYourSubs is a directory of apps that support it, so we use the term every day.

What does bring your own subscription mean?

An app supports BYOS when you can sign in with, or connect, a consumer AI plan and the app's AI requests are then counted against that plan. Raycast's manual has a clean one-line version: "Bring Your Own Subscription lets you use your Claude or ChatGPT subscription inside Raycast," and "Usage counts towards that subscription, not your Raycast AI credits."

Three things are true of every BYOS setup:

1. You already pay the AI provider. The plan is yours, in your name.
2. The app does not resell AI to you. It may charge for its own features, but not for the model calls.
3. Your plan's limits apply. Claude and ChatGPT plans have a rolling five-hour window and a weekly cap, and usage in every connected app is shared with your own chats.

How is BYOS different from BYOK (bring your own key)?

BYOK came first. You paste a developer API key into an app, and the provider bills each token to that key. BYOS swaps the key for your consumer plan.

	Bring your own subscription (BYOS)	Bring your own key (BYOK)
What you connect	Your Claude or ChatGPT plan (sign-in)	A developer API key from a provider console
How you pay	Flat monthly plan price	Per token, at API list prices
What limits you	The plan's five-hour and weekly limits	Your spend and the key's rate limits
Bill shock risk	None on the model, unless you turn on paid credits	Real; Raycast tells BYOK users to "monitor the provider's API console to prevent costly surprises"
Best for	Heavy personal use, agents, coding	Shared automation, CI, production workloads
Who it is for	People with a Pro, Max, Plus or Pro plan	Anyone with a provider billing account

Raycast's BYOK page puts the BYOK side plainly: "You'll be responsible for the API costs incurred at the provider's standard rates." Its BYOS page adds that "accounts that only have API access don't qualify" for BYOS. For a worked comparison of what a plan saves against API billing, see our Claude subscription vs API cost guide.

How does BYOS work with ChatGPT plans?

OpenAI supports it directly, in two ways.

• Sign in with ChatGPT. OpenAI's partner page says that in supported apps "eligible ChatGPT Plus and Pro subscribers can also choose to use their ChatGPT plan for AI requests." You approve a separate "Use your ChatGPT plan" permission, and in ChatGPT Settings, Usage you can cap each app at a percentage of your weekly usage. OpenAI lists Amp, Conductor, Dactyl, Devin, Hermes Agent, Hyperagent, Kilo Code, Notion, Vercel (v0), Vorflux and Warp among the partners, plus open-source integrations OpenClaw, OpenCode, Pi and T3 Code.
• Codex sign-in. Many open-source apps sign you in to Codex with your ChatGPT account (browser OAuth or a device code) or drive the Codex CLI you are already signed in to. OpenClaw's authentication page compares the two routes side by side: both use "your Codex allowance."

OpenAI is explicit about cost: "App usage counts toward your existing plan limits. Connecting an app does not add a new allowance, and the app may have its own fees." It also says using your plan "does not give the app access to your ChatGPT conversations or memories."

The ChatGPT hub lists every app on BringYourSubs that supports either route.

How does BYOS work with Claude plans?

More narrowly. Anthropic's Claude Code legal page allows one route: an app may run the unmodified Claude Code, and you sign in to it with your own Claude account. Its wording: the restrictions do not "prevent an end user from signing in to the unmodified Claude Code binary with their own Claude subscription, including where a platform hosts Claude Code."

The same page rules out the other routes: "Anthropic does not permit third-party developers to offer Claude.ai login into their own applications, or to route requests through Free, Pro, or Max plan credentials on behalf of their users," and developers "may not collect, store, or intermediate Claude.ai credentials or session tokens."

In practice, Claude BYOS apps ask you to install Claude Code and log in to it, then they drive it. Cline, Zed, Conductor, Happy and OpenClaw (through its "Claude CLI" route) work this way. Raycast does too: its manual says you sign in by running claude and /login. Anthropic's usage limits page confirms the shared meter: "your usage of all different Claude product surfaces (claude.ai, Claude Code, Claude Desktop) counts towards the same usage limit."

Our guide Can I use my Claude subscription in other apps? goes app by app, and the Claude hub lists the current apps.

What about Cursor plans?

"Cursor bring your own subscription" is a real search, and the answer is narrower again. Cursor's plan pays for Cursor's own editor and its CLI agent. A handful of apps, such as T3 Code, Superset and Munder Difflin, drive that CLI so its work runs on your Cursor plan. We know of no supported way for an app to use a Cursor plan as a general route to Claude or GPT models. The Cursor hub has the list, and our post on using a ChatGPT subscription in Cursor covers the reverse question.

Why are people switching to BYOS?

Three reasons stand out:

• One bill instead of several. If you pay $20 or more a month for Claude or ChatGPT, paying a second AI subscription inside each app duplicates the cost.
• No metered surprises. A flat plan caps your spend. Agents that loop for an hour can run up real API bills; on a plan they hit a limit instead.
• Better models by default. Plans include the provider's top models. On ChatGPT Plus, for example, OpenAI's Codex pricing page lists GPT-6.1 Sol and GPT-6 Luna.

The trade-off is the limit itself. OpenAI's Plus estimates for one five-hour window range from 5 to 45 local messages on GPT-6 Astra to 350 to 3,000 on GPT-6 Luna, and weekly limits may also apply. Heavy agent use can run a plan dry; that is when BYOK, or paid credits, makes sense.

What should you check before connecting your plan to an app?

1. Which route does it use? For ChatGPT: Sign in with ChatGPT or Codex sign-in. For Claude: the official Claude Code you log in to. Be wary of a Claude app that asks for your Claude password, a pasted token, or its own "Sign in with Claude" screen.
2. Is there a per-app cap? ChatGPT lets you set one in Settings, Usage. Claude does not have per-app caps, so watch Settings, Usage yourself.
3. What happens at the limit? Some apps fall back to an API key or their own credits; Raycast says it "doesn't switch you to Raycast AI credits automatically."
4. What does the app charge for itself? BYOS removes the model bill, not the app's own price.
5. Can you disconnect it? For ChatGPT, go to Settings, Security and login, Sign in with ChatGPT. For Claude, sign out of Claude Code with /logout.

Our step-by-step guide on running other apps on your Claude or ChatGPT plan shows how to confirm an app is billing your plan and not an API key, and the ranked list of the best agent apps on your subscription is a good place to start.