Claude Code's June 15 Billing Change: Keep Remote Sessions on Your Subscription

Claude Code's June 15 Billing Change: Keep Remote Sessions on Your Subscription

Starting June 15, 2026, Anthropic changes how Claude Code is billed on Pro, Max, Team, and Enterprise plans. If you use Claude Code remotely — from your phone, a web dashboard, or any automation — this change probably affects you. Here is what's changing, who pays more, and how to keep your remote Claude Code sessions covered by the subscription you already pay for.

What changes on June 15, 2026

According to Anthropic's official announcement:

  • Claude Agent SDK and claude -p (headless / non-interactive mode) no longer count toward your plan's usage limits.
  • Instead, eligible plans receive a separate, much smaller monthly Agent SDK credit:
Plan Monthly Agent SDK credit
Pro $20
Max 5x $100
Max 20x $200
Team (Standard seat) $20
Team (Premium seat) $100
  • The credit covers Agent SDK projects, claude -p runs, the GitHub Actions integration, and third-party apps that drive Claude through the Agent SDK. Once it's used up, you pay metered API rates.
  • Interactive Claude Code sessions are not affected. Your plan's session and weekly limits stay reserved for interactive use — the same as typing in your terminal.

Who gets hit by this

The dividing line is interactive vs. non-interactive, not local vs. remote. The problem: most tools that let you control Claude Code remotely are built on headless mode (claude -p or the Agent SDK), because it's the easiest way to integrate. From June 15, every message you send through such a tool burns the small SDK credit instead of your subscription — and a single heavy day of agent work can cost more than the entire monthly credit. There are already reports of subscribers billed over $1,800 in two days of claude -p usage.

So if you:

  • check on long-running Claude Code tasks from your phone,
  • drive Claude Code from another device while away from your desk, or
  • use a third-party remote client built on the Agent SDK,

…you should check which mode your tool uses before June 15.

How PandaNpc keeps remote sessions on your subscription

PandaNpc takes the other path. Its Interactive Mode doesn't spawn claude -p and doesn't use the Agent SDK. Instead, the open-source pandapaw bridge runs the real interactive Claude Code session on your own computer — the exact same REPL you'd get by typing claude in your terminal — and mirrors it to your phone, browser, or another desktop:

  • Billing-wise it is an interactive session. Usage draws from your plan's normal interactive limits (the 5-hour and weekly windows), not the $100–$200 SDK credit pool.
  • Full remote control: send messages, approve or deny tool permissions, answer Claude's questions, switch models, interrupt generation — from an iPhone app, the web, or the desktop client.
  • Live status HUD: see your 5-hour / weekly usage windows, context size, and an estimated cost equivalent in real time, so you always know where your subscription limits stand.
  • Sessions survive disconnects: close your phone, reopen later, and pick up the same session with full history.

PandaNpc also supports a Headless Mode for users who want SDK-credit-backed background automation — you choose per connection. But for day-to-day "drive Claude Code from anywhere" use, Interactive Mode means the June 15 change costs you nothing extra.

Set it up in two minutes

  1. On the computer where Claude Code runs, install the bridge:
curl -fsSL https://cos.pandanpc.com/pandapaw/install.sh | sh

(Windows: iwr -useb https://cos.pandanpc.com/pandapaw/install.ps1 | iex)

  1. The installer prints a QR code. Scan it with the PandaNpc iOS app — or open the web client / desktop app and add the connection.
  2. Start or resume a Claude Code session from any device. That's it — no API key, no SDK credit consumed.

See the PandaPaw setup guide for details.

FAQ

Does PandaNpc's Interactive Mode consume the new Agent SDK credit? No. It attaches to a genuine interactive Claude Code session on your machine. Anthropic's announcement explicitly keeps interactive Claude Code on your plan's regular usage limits.

Do I need a Claude subscription? Interactive Mode works with whatever your local claude CLI is signed into — Pro, Max 5x, or Max 20x. The same limits you'd have in the terminal apply remotely.

What about my CI pipelines and cron jobs? Truly unattended automation (no human in the loop) is exactly what the Agent SDK credit is for. Use PandaNpc's Headless Mode or plain claude -p for those, and keep the interactive windows for your hands-on work.