Playbooks · Coding tasks
Run Claude Code Projects without burning quota
Written November 16, 2026
Claude Code Projects run every thread as its own session, new projects default to Opus at high effort, and PR-watching threads spend again.
A Claude Code Project is one ongoing conversation where Claude starts a thread for each task and runs those threads in parallel. That shape is the point of the feature, and it is also why the plan limit arrives sooner than it does in a single session. Anthropic's Projects documentation states the trade plainly. Project usage counts against the same plan limits as other Claude Code sessions, and a project uses those limits faster than a single session does. On a Pro plan in particular, you should expect to reach the limit sooner on days you run one. Projects are in public beta on Pro, Max, and Ultra plans. The controls below all come from that same page.
What draws on your plan
Three things spend, and only one of them is obvious:
- Running threads. Each thread is a full session, and several can run at once. There is no fixed number: Claude starts as many as the work calls for. A limit you ask for is a preference rather than a cap. The enforced limit, per the Projects documentation, is 500 new threads per day across your projects.
- The conversation itself. Claude uses tokens of its own reading what threads report and deciding what to do next, before any thread does new work.
- Threads watching a pull request. An idle thread wakes up and uses your plan again when CI fails or a review comment arrives on its pull request. To stop that, ask in the thread for it to stop watching the pull request.
The other side of that list is useful too. A project with no running threads, no watched pull requests and no new messages does not use your plan while it sits idle, and neither does an archived project.
The defaults that keep the spend going
Model and effort. A new project runs every thread on Opus at high effort, which the documentation says draws on your plan fastest. Thread model and thread effort are set in Project settings, under General. The conversation runs at low effort by default, with its own coordinator model and effort settings in the same place.
Auto-continue. A thread that stops at the plan limit waits, and by default continues on its own when the limit resets. Its work then counts against the next usage window. The setting is called Auto-continue when usage limits reset, it sits in Project settings under General, it is on by default, and it applies to the whole project rather than to one thread. Turn it off there, before the reset time the thread's message shows, if you do not want work to restart on its own.
Past the limit. Work goes past the plan limits only if you have turned on usage credits for your account. A thread cannot turn them on for you, and a project cannot spend past the limits on its own.
See the spend, then bring it down
Open Usage in Project settings to see token use by thread and by model, and how much went to the project conversation. That view is where a quota problem stops being a guess: it shows which thread and which model did the spending.
- Mind the one-hour point. On Pro, Max, and Ultra, within the plan limits, the cache lifetime is an hour. A thread idle longer than that rereads its whole conversation before doing anything when a follow-up reaches it. For new work, ask Claude to start a fresh thread, which can use less than reviving a large old one.
- Use a smaller model or lower effort for work that does not need the largest model — for threads, the conversation, or both.
- Ask for fewer threads at a time, or ask Claude in the project conversation to answer small questions itself instead of starting a thread.
- Stop pull request watching in any thread that no longer needs to react to CI failures or review comments.
When you need everything to stop
Pause and Archive both sit at the bottom of Project settings, under General. Pause stops everything at once: every running thread and the conversation are interrupted, no new threads start, scheduled tasks do not run, and the project accepts no messages until you resume it. Archive hides the project from the sidebar and archives its threads, which stops any thread that was running or watching a pull request; scheduled tasks in the project do not run while it is archived. Either one is a firmer stop than asking a busy project to slow down, because an instruction to Claude about thread counts is a preference, not a cap.
Related playbooks
- Run Sonnet with an Opus advisor in Claude Code
- Run claude-mem on your own plan, redaction on
- Run LongCat-2.5 in Claude Code without the 200K cap
- Run three independent lines of research at the same time in Claude Code, then combine the short summaries
Sources
- Let Claude coordinate ongoing work with Projects — Claude Code Docs — Usage and cost, the 500 new threads per day enforced limit, the Opus and high effort default, auto-continue, the Usage view, the one-hour cache lifetime, and Pause and Archive in the Project settings reference.
Back to: Coding playbooks hub · Also: Claude Code 2.1.317: schedule warnings and cost reports