Anthropic just repriced agent usage. If you run Claude through Zed, Conductor, T3 Code, superset.sh, or anything else built on the Agent SDK, your Max 5x plan now covers a lot less than it did last week. Lanes does not change, because Lanes was never built on the part that just got repriced. Here is what changed, who it hits, and why your Lanes sessions keep drawing from the same subscription pool they always did.
What changed
On May 14, Anthropic emailed plan subscribers about a new monthly Agent SDK credit, effective June 15. The full policy is in Anthropic's support article on the new Agent SDK credit:
Starting June 15, 2026, Claude Agent SDK and
claude -pusage no longer counts toward your Claude plan's usage limits. Your subscription usage limits stay the same and stay reserved for interactive use of Claude Code, Claude Cowork, and Claude.
To cover Agent SDK usage, every eligible plan now gets a separate monthly credit:
- Pro: $20 / month
- Max 5x: $100 / month
- Max 20x: $200 / month
- Team Standard seats: $20, Team Premium seats: $100
- Enterprise (usage-based): $20, Enterprise (seat-based Premium): $200
The credit is per-user, refreshes monthly, does not roll over, and drains first. When it runs out, additional Agent SDK usage routes to extra usage at standard API rates, but only if you have explicitly enabled extra usage. Otherwise your SDK requests stop until next cycle.
Anthropic is also clear about the credit's intent:
The Agent SDK monthly credit is sized for individual experimentation and automation. Teams running shared production automation should use the Claude Developer Platform with an API key for predictable pay-as-you-go billing.
If you are running multiple agents in parallel through a tool that drives the SDK, that paragraph is talking about you. The credit was not sized for it.
The dividing line is which side of the SDK boundary your workflow sits on. Anthropic lists the covered surface as: the Agent SDK in your own projects, claude -p in non-interactive mode, the Claude Code GitHub Actions integration, and "third-party apps that authenticate with your Claude subscription through the Agent SDK." What stays on your existing subscription is:
- Interactive Claude Code in the terminal or IDE
- Claude conversations on the web, desktop, or mobile apps
- Claude Cowork
That second list is the one that decides which side of the bill your workflow lands on.
For a sense of the scale of the shift, Zed's own post on the change notes that subscriptions previously subsidised agent usage at roughly 15 to 30x compared to API pricing. The new credit is billed at full API rates.
Who this hits
If your workflow drives Claude programmatically, you are in the new pool.
Zed runs Claude through the Agent Client Protocol, and has confirmed in their own post that any usage going through ACP draws from the credit. T3 Code, Theo Browne's open-source control plane, drives Claude through the same kind of SDK surface, and Theo has been publicly vocal about a roughly 40x reduction in effective subscription value. Conductor and superset.sh are the parallel-agent orchestrators most commonly named alongside Zed in this discussion, both flagged by the developer community as affected. Claude Code GitHub Actions, OpenClaw, and anything else built on @anthropic-ai/claude-agent-sdk are in the same boat.
Zed has already published the workaround in their own words: "If you want to keep using Claude with your Claude Pro or Max plan at your existing subscription limits, you can run Anthropic's official claude CLI in a terminal inside Zed instead of through ACP. When the official claude CLI runs in the terminal, it uses your subscription's limits, not the new credit."
That workaround is not a workaround in Lanes. It is the default path. It is the only path.
Why Lanes is not in the new pool
Lanes does not wrap the Agent SDK. Lanes does not call claude -p. Lanes does not implement ACP. When you start a session on an issue, Lanes opens a real pseudo-terminal (using a native PTY library), boots your real login shell with your real dotfiles, and runs the official claude command exactly as you would if you typed it yourself.
Same binary. Same flags. Same authentication. Same subscription pool.
From Anthropic's billing perspective, a Lanes session is indistinguishable from a developer sitting at a terminal and running claude --resume <session-id>. Because that is what it is. The exact phrase from Anthropic's own support article ("Interactive Claude Code in the terminal or IDE") describes the surface every Lanes session runs on.
What you keep when you switch
Everything you came to your current tool for, mapped onto an architecture that did not move.
Parallel orchestration and oversight. Each agent gets its own issue card on a board you drag through Planning, Implementation, Review, and Done. Each card has a live terminal with real-time status: busy, awaiting input, stopped, error. The bell rings when an agent needs you. The same parallel-agent surface Conductor and superset.sh sell, except the runtime underneath is the native CLI.
Isolated worktrees per session. Lanes auto-creates a git worktree with a generated branch name for each issue, tracks uncommitted and unmerged state in real time, and cleans up when the issue lands in Done. No collisions, no stomping, no hand-juggling of worktree paths.
Automation. Linear and GitHub integrations pull tickets straight onto the board and write PR links back. Quick commands inject preset prompts into any session.
All of it sits on top of the interactive pool. Your Max plan keeps doing what it did last week. Your fleet of agents keeps running. No metering on the programmatic side, no $100 ceiling, no surprise burndown halfway through a sprint.
Where this leaves us
Software is in a pricing-discovery phase. Anthropic is the first to make the split explicit, and they will not be the last. Architectures that route through the first-party CLI keep working through these shifts. Architectures that wrap the SDK keep paying the re-pricing tax every time the policy moves.
Lanes was built the first way on purpose.
If you have been running multiple agents through a tool that just got more expensive, start here or join our Discord. Bring your Max plan with you.