How Fluf runs Claude Code with your own login (2026 rules)

You already pay for Claude. Maybe it's Pro at $20 a month, maybe it's Max because you live in Claude Code. Now you want a team of bots doing real work: research, inbox triage, drafts, code, the boring admin. The obvious question is whether that plan can power them, or whether you need a second AI bill on top.

Short answer: yes, your plan can power them through Anthropic's Claude Code, installed and signed in on your computer, because that is the way Anthropic allows. In 2026 that line moved several times, broke a lot of popular tools, and filled timelines with confusing headlines. This article walks through what happened, quotes the rules as they stand today, and gives you a checklist for any app that promises to "use your Claude subscription".

Fluf is not affiliated with, endorsed by or partnered with Anthropic. Claude and Claude Code are Anthropic's products.

What happened in 2026, in order

If you only skimmed the news, it looked like Anthropic banned everything. It didn't. Here is the sequence as reported.

• 9 January. Server-side checks started blocking subscription logins outside Claude Code. Tools such as OpenCode, Cline, RooCode and OpenClaw stopped working with Claude plans. According to this ecosystem write-up, several of them had been sending a request header that made them look like Claude Code.
• 17 to 20 February. Anthropic updated its terms to put the rule in writing (Gigazine, AlternativeTo).
• 19 March. Anthropic reportedly sent legal demands to OpenCode, which then removed its Anthropic login (same write-up).
• 4 April. Claude Code lead Boris Cherny said Claude subscriptions would "no longer cover usage on third-party tools like OpenClaw", citing strain on capacity. Affected users were offered a credit or refund (Cybersecurity News).
• 15 June. Anthropic paused a previously announced change to how Agent SDK usage is billed. Its support page says Agent SDK, claude -p and third-party app usage still count against your plan's limits (Claude support).
• 7 October. The same support page added that Max and Team plans now include monthly API credits, and states: "You can still use the Claude Agent SDK, claude -p, and third-party apps with your subscription limits." A secondary report puts those credits at $100 a month for Max 5x and $200 for Max 20x. We couldn't confirm the amounts on an Anthropic page, so treat them as reported.

The April statement and the October support page read differently, and the policy has clearly shifted more than once. That's the honest state of things. The fix is not to guess from headlines. It's to read what Anthropic publishes today.

What the rules actually say today

Anthropic's Claude Code legal and compliance page is the clearest source. Three passages matter.

On how sign-in works:

"OAuth authentication is intended exclusively for purchasers of Claude Free, Pro, Max, Team, and Enterprise subscription plans and is designed to support ordinary use of Claude Code and other native Anthropic applications."

On what third-party developers can't do:

"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. Moreover, developers may not collect, store, or intermediate Claude.ai credentials or session tokens — sign-in to a Claude account must complete through Anthropic's own flow."

And the line that makes bots on your own plan possible:

"Nor does it 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 sets conditions for any product that runs Claude Code: "The Claude Code binary must not be modified," and products "may not pay for, resell, or intermediate Claude usage on their end users' behalf." Each user authenticates with their own plan, their own API key or their own cloud provider account, and is billed directly.

Put plainly:

• Allowed: an app runs the official Claude Code program, untouched, on your machine. You sign in to it yourself through Anthropic's own sign-in. Usage lands on your plan.
• Not allowed: an app shows its own "Log in with Claude" box, asks for your Claude password or session token, stores it, fakes being Claude Code, or sells you Claude usage it buys in bulk.

One more rule sits on top: Anthropic "reserves the right to take measures to enforce these restrictions and may do so without prior notice." Tools that cut corners can break overnight. January proved it.

The limits you're working with

Signing in the right way doesn't make your plan unlimited. According to Claude's pricing page, "Every plan has usage limits that reset on a rolling five-hour session window, and paid plans add weekly limits on top." Pro gives "at least 5x more usage per 5-hour session than Free". Max comes in 5x and 20x tiers over Pro, priced "from $100" a month. Anthropic doesn't publish exact caps.

The legal page adds a key phrase: "Advertised usage limits for Pro and Max plans assume ordinary, individual usage of Claude Code and the Agent SDK." That's the spirit to respect. Your plan is for you. It is not a back end for a company of fifty people, and it isn't something to resell.

Limits are also where people get burned in practice. In 2026, users reported Max sessions running out much faster than expected, and a September change to weekly Claude Code capacity drew pushback (DevOps.com, Implicator). Anthropic framed the September figure as an increase over pre-promotion levels, so read both sides before drawing conclusions. Either way, the lesson for anyone running bots is simple: you need to see your remaining usage before a bot hits the wall in the middle of a job.

When to add an API key

A subscription is the cheapest way in for one person doing focused work. An API key makes sense when:

• You're building a product for other people. Anthropic is explicit that developers building on the Agent SDK should use API keys.
• You want bots running long, heavy jobs. Pay-per-token billing has no five-hour window to wait out.
• You want a fallback. When your plan's window runs dry, a bot can keep going on a key instead of stalling.
• You've got credits to use. If you're on Max or Team, the monthly API credits Anthropic announced in October can go toward this.

Most solo operators end up with both: the plan for day-to-day work, a key as a safety net.

How Fluf does it

Fluf is a desktop app (Mac and Windows today; Linux is in release testing) where you run a team of AI bots, each with a name, a face and a job, organised like a real company. Here's exactly how Claude fits in.

• The official Claude Code app does the Claude work. Fluf runs Anthropic's unmodified Claude Code program on your computer. We don't patch it, wrap its sign-in or pretend to be it.
• You sign in through Anthropic. When you connect Claude, Anthropic's own sign-in opens. Your credentials go to Anthropic, not to us.
• Fluf never sees or stores your Claude credentials. No password box, no token field, no copy on our servers. Fluf's account service only handles your Fluf account, licence, billing and downloads.
• Usage counts against your plan. We don't buy Claude usage and resell it. Every token your bots use is billed by Anthropic to you, under your agreement with them.
• Remaining usage is right there. Settings › AI connections shows each connection's remaining plan usage and how many tokens each bot has used, so you can spot a heavy bot before it drains your window.
• An API key is one click away. Add your own Anthropic API key as a fallback, or move a specific bot onto it.
• Several accounts, kept separate. You can connect more than one Claude account and rename each one, for example your personal plan, your work plan, or a client's plan on a client's business. That's about keeping separate people and workspaces separate. It is not a way to pool accounts to get around limits, and we don't design it to be.

The same approach covers other providers: your installed Codex CLI with your own login, plus API keys for Anthropic, OpenAI, Google, xAI, OpenRouter and others. Fluf itself costs a flat fee: Free covers 2 businesses and 6 bots, Pro is $19 a month or $139 a year. Your model usage is between you and your provider. No credits, no markup.

A checklist for any tool that says "use your Claude plan"

Before you connect your subscription to anything, ask five questions.

1. Does sign-in happen on Anthropic's own page? If the app asks you to paste a token or type your Claude password into its own form, walk away.
2. Does it run the official Claude Code program, unmodified? If it imitates Claude Code instead of running it, it can break without warning, and your account may be the one affected.
3. Who pays Anthropic? It should be you, directly. A tool selling you "Claude credits" on top of your plan is in territory the terms rule out.
4. Can you see your remaining usage? Bots that work while you're busy elsewhere need a visible fuel gauge.
5. Is there an API key fallback? Limits will be hit. A good tool keeps working instead of failing silently.

A tool that passes all five is using your subscription the way Anthropic describes. A tool that fails any of them is a risk, however good the demo looks.

The bottom line

You can run a team of AI bots on the Claude plan you already pay for, through Claude Code. The rule is simple once you read it: the official Claude Code app, your own sign-in through Anthropic, usage billed to you, ordinary individual use. The tools that broke in 2026 broke because they cut corners on that, not because the idea was wrong.

---

Give your Claude work a whole team. Fluf gives you bots with names, faces and an org chart, running on your own computer. Fluf runs your installed Claude Code, signed in through Anthropic's own flow, add any other model you like, and watch every bot's usage live. Start free with 2 businesses and 6 bots, or try Pro free for 14 days. No card needed.

Download Fluf

Read next: How to build an AI org chart · AI employees without credit burn