Anthropic raised Claude Code weekly usage limits by 50%, effective immediately and running through July 13 at 6PM PDT (1AM GMT, July 14). It is live on every paid plan (Pro, Max, Team, and seat-based Enterprise) and applies everywhere Claude Code runs: CLI, IDE extensions, the desktop app, and the web. There is nothing to opt into. The new ceiling is already on your account.
The interesting part is what this stacks on top of. Last week, Anthropic doubled the 5-hour limits. Combined, that is a 2x boost to your short-window quota and a 1.5x boost to your weekly cap, all live at the same time, all without a price increase. For two months you get a Claude Code account that runs significantly hotter than it did in April.
This post breaks down what the limits actually are, what the 50% increase looks like in practical terms, what the smart play is for the next eight weeks, and how to plug the extra capacity into real engineering work, including API design with Apidog.
What the weekly limits are, and what 50% more means
Claude Code uses a two-tier rate system on paid plans:
- 5-hour limits. A rolling cap on tokens you can consume in any 5-hour window. This is the limit you hit during a long single session, like a multi-hour debug or a feature build. Anthropic doubled this last week.
- Weekly limits. A cap on total tokens across the full week. This is the one that quietly catches power users at the end of a heavy sprint. This is what just went up 50%.
The exact token numbers depend on your plan tier and are not published as fixed values; Anthropic reserves the right to tune them based on capacity. Roughly, the new ceiling means:
- Pro users get noticeably more headroom before they hit the warning that throttles output until the week rolls over.
- Max users, who already had the highest individual ceiling, can now sustain a multi-day deep work block on a single complex codebase without hitting a wall.
- Team and seat-based Enterprise users get the same 50% lift per seat. For a 10-seat Team, that is a meaningful weekly capacity increase across the org.
You will see the change in the same place you always saw your usage: the /usage command in the Claude Code CLI, the status bar in the IDE extensions, and the account settings page on the web.
Why Anthropic is doing this now
Two reasons worth naming.
First, capacity. Anthropic has been provisioning new compute aggressively through Q1 and Q2 of 2026. When you have spare capacity, the cheapest way to deepen platform loyalty is to give existing paying users more of what they are already paying for. It also flushes out usage patterns that inform future pricing tiers.
Second, competition. Codex shipped its own /goal autonomous agent loop a few weeks ago, and OpenAI has been quietly bumping rate limits on its developer plans too. The 50% lift through July 13 is a defensive move against engineering teams who were starting to split their workloads across vendors.
Worth noting: the temporary window matters. July 13 at 6PM PDT is when the boost ends. Anthropic has not said whether the new ceiling will become permanent, drop back to baseline, or land somewhere in between. Plan accordingly.
What the 50% increase actually unlocks
Generic “more usage” coverage misses the point. Here is what the extra weekly quota maps to in practice.
Longer-horizon agent runs
The combined effect of doubled 5-hour limits and 50% more weekly capacity is that the /goal autonomous loop can now run for substantially longer before you hit a ceiling. Before this change, a long /goal run on a complex refactor would eat through your 5-hour budget and force a break. With both limits up, you can hand Claude Code a goal with measurable success criteria and let it run end-to-end without babysitting.
This is the single biggest practical unlock. Long-horizon agent work was rate-limit constrained, not capability constrained, for a lot of Pro and Max users.
Bigger codebase context
Claude Code’s value scales with how much of your codebase it can hold in context. With more headroom on both windows, you can load larger working sets, run agents across deeper directory trees, and skip the manual scoping that used to be necessary to stay under the cap. If you have a monorepo that previously felt “too big for Claude Code,” the next two months are when to test that assumption again.
Multi-agent workflows
Tools like Ruflo, the multi-agent orchestrator on top of Claude Code, spin up several Claude instances against the same task and merge their work. These workflows previously ate quota fast because you were running 3-5x the calls in parallel. With the new limits, they become much more practical for daily use rather than special-occasion experiments.
Heavier MCP server usage
Claude Code calls external tools through the MCP (Model Context Protocol), and those calls count against your quota. If you have been holding back on chaining MCP servers (database access, API testing, browser automation, GitHub triage), now is the time to wire them up. Every MCP tool call ate a chunk of your old weekly budget; under the new ceiling, you can build a workflow that uses 5 or 6 MCP servers in series without the math feeling tight.
If you are running into config errors that block MCP server setup in Claude Code, the invalid custom3p enterprise config fix covers the most common cause.
The smart play for the next eight weeks
You have until July 13 to take advantage of the lift. Three concrete moves:
- Ship the agent workflow you have been postponing. If you had a
/goal-based or multi-agent setup that was hovering at the edge of viable on the old quota, build it now. Even if the limits drop back to baseline after July 13, you will have proven out the workflow and can decide whether to upgrade plans to keep it. - Move side-channel work into Claude Code. Tasks you were doing in cheaper tools or by hand (writing AGENTS.md files, code review, MCP server configuration, OpenAPI spec generation) can move into your main Claude Code session without blowing the weekly budget. See our guide to writing AGENTS.md files for one of the easier wins.
- Test the new ceiling honestly. Run Claude Code the way you would if the new limits were permanent. If you find you can comfortably sustain your daily work under the new ceiling, you have data for the decision you will face in July: stay on Pro, upgrade to Max, or split your workload across providers.
Where API work fits in
For backend and platform engineers, the highest-leverage use of the extra quota is API work. Claude Code is exceptionally good at API tasks (writing handlers, generating OpenAPI specs, debugging contract mismatches, building integration tests) but those tasks tend to be the ones that eat the most tokens because they require deep file context and frequent tool calls.
A workflow that benefits directly from the new limits:
- Design the API contract in Apidog. Define the endpoints, request and response schemas, and example payloads. This is your source of truth.
- Export the OpenAPI spec and hand it to Claude Code as context.
- Use
/goalto implement against the spec. Tell the agent to keep iterating until every endpoint passes its Apidog test cases. With the new limits, this loop can run end-to-end without you having to pause and resume across multiple sessions. - Run the Apidog CLI tests as the validator. Each loop iteration checks the real contract, not a fake one the agent invented.
The deeper guide on contract-first API work with AI agents lives at our design-first API workflow guide. The point is that you finally have enough weekly quota to run agents against real API work end-to-end, instead of partial sessions that you have to manually stitch together.
If you have not used Apidog before, download Apidog and try the design-first workflow with the extra Claude Code quota you just got.
What about free Claude API access?
For people who do not want to pay for a Claude Code plan, the free Claude API access guide covers the paths Anthropic and partners offer. Those are separate from Claude Code’s paid quota and are not affected by this 50% increase. The 50% lift only applies to Pro, Max, Team, and seat-based Enterprise.
A note on what this does not change
To set expectations correctly, this update does not:
- Change Claude API rate limits for direct API users (those are governed by your Anthropic API account, not your Claude Code plan)
- Change pricing on any tier
- Add new features to Claude Code itself
- Affect billing for enterprise seat-based plans (the same per-seat structure applies)
It is a usage cap lift, nothing more, nothing less. The value is in what you do with the extra room.
FAQ
When exactly does the 50% increase end? July 13, 2026 at 6PM PDT (1AM GMT, July 14). After that, Anthropic has not announced what the ceiling becomes. Assume it reverts to baseline unless they say otherwise.
Do I need to do anything to enable it? No. The limits are already raised on your account. Check the /usage command in the Claude Code CLI to see the new ceiling reflected in your status output.
Does the 50% increase apply to the 5-hour limit too? The 50% specifically applies to the weekly limit. The 5-hour limit was separately doubled (2x) the week before. Both are live concurrently through July 13.
What if I upgrade to a different plan during the window? Anthropic has not published guidance on this specifically. Based on how their usage windows have worked historically, the new tier’s limits apply at the new (already-raised) levels from the moment you upgrade.
Does this affect Claude Code on the Anthropic API plans? No. This change is to Claude Code plans only (Pro, Max, Team, seat-based Enterprise). If you call Claude through the Anthropic API directly, your rate limits are governed separately and have not changed.
Will this become permanent? Unknown. Anthropic explicitly framed it as a window through July 13. Treat the next eight weeks as a chance to test whether you actually need the higher ceiling, then decide based on real usage data.
