Quick wins for a faster PC:
Scan for outdated or missing drivers - takes under a minuteDriver Scan →Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →You can reduce Claude Code costs without losing the information a task depends on by measuring usage, dropping stale context between unrelated tasks, and deliberately preserving the right details when you compact an ongoing session. Then match the model and tools to the work instead of paying for unnecessary reasoning or repeated output. The best changes depend on your account type, model, codebase, and habits, so verify them against your own usage.
Measure usage before changing your workflow
Start with Claude Code’s /usage command. It shows token statistics for the current session and can help identify whether context, output, or a particular work pattern is driving use. For API users, it also shows an estimated dollar figure based on list prices unless organization-managed pricing is configured. Anthropic says the Claude Console Usage page is authoritative for API billing; the session estimate is not a substitute for the billing record.
![Claude AI for Beginners Bible: [5 in 1] The Ultimate Guide to Automate Your Work, Save Hours Every Week, and Use AI for Real-World Results](https://m.media-amazon.com/images/I/415+fSJacsL._SL160_.jpg)
Pro and Max subscribers see plan usage information. The API-style session cost estimate is not their subscription bill, so do not treat it as an invoice or use it alone to judge the value of a plan. The appropriate place to check actual usage or charges depends on how you access Claude Code: subscription, Console API, or a cloud-provider deployment. See Anthropic’s Claude Code cost guidance for the account-specific reporting details.
Record a representative task before changing settings, then compare similar work afterward. A small bug fix and a large refactor are not comparable tests: codebase size, model, task difficulty, and the amount of output can all change usage. There is no single cost target that applies to every developer or team.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problems#1 Best Overall
Choose whether to clear or compact a session
The key distinction is whether the next task needs the current conversation. Anthropic summarizes the relationship this way: “Token costs scale with context size: the more context Claude processes, the more tokens you use.” The practical goal is not to minimize context indiscriminately; it is to keep the context that helps Claude act correctly and remove material that no longer matters.
Use /clear for unrelated work
When switching to a separate task, use /clear so old discussion and investigation do not continue consuming attention in the new session. If you may need to return to the earlier work, rename that session before clearing or leaving it, so it is easier to find and resume.
Use /compact when work continues
For a task that is still underway, compact the conversation rather than discarding its useful history. Give /compact instructions about what must survive—for example, decisions made, relevant code changes, test output, or API details. A generic summary may omit precisely the result or constraint the next step needs.
For recurring project-specific requirements, add concise compact instructions to CLAUDE.md. Keep them focused on durable information; a long file of instructions that is irrelevant to most tasks can itself become needless context.
Match the model and reasoning to the task
Do not default to the most capable model for every request. Anthropic’s cost guide says Sonnet handles most coding tasks at lower cost than Opus, and recommends reserving Opus for demanding architectural decisions or multi-step reasoning. It also suggests Haiku for simple subagent tasks. Model availability and rates can change, so check the current Anthropic pricing documentation before relying on a specific comparison.
Use the least costly model that can reliably do the job, and escalate when the task warrants it. A narrow, well-specified edit is different from evaluating architectural trade-offs across a codebase. For team workflows, a simple subtask assigned to a subagent may not need the same model as the main reasoning task.
Reasoning effort is another choice to make deliberately. Anthropic documents thinking tokens as billed output tokens; lowering effort can reduce token use on straightforward work where extended reasoning is unnecessary. Keep greater effort for tasks that benefit from it. Controls differ among model families, and some models use thinking that is always on, so do not assume one setting behaves identically everywhere. Anthropic’s prompting guidance also discusses lower effort when overthinking is undesirable.
Cut tool and output overhead without hiding important results
- Disable MCP servers you are not using. Unneeded server tools can add definitions to the context Claude has to process. Turn them back on when a task needs them.
- Prefer a CLI where it avoids tool-list overhead. If a command-line tool can perform the needed operation without exposing a large set of unused MCP tools, it may keep the available tool context smaller.
- Inspect context with
/context. Use it to see what is occupying context rather than guessing which integration or instruction is responsible. - Filter oversized command output. Hooks can filter large outputs before Claude receives them. Preserve the errors, changed files, or other evidence needed for the task, and remove bulk that is not useful.
- Put specialized knowledge in skills. Skills can supply workflow-specific information when needed, instead of making every instruction persist in every task.
- Keep
CLAUDE.mdessential. Put durable project guidance there, and move instructions that apply only to a particular workflow into a skill.
Reducing context is not a win if it removes test failures, relevant code, or tool results Claude needs. Use /context to identify overhead and filter selectively rather than suppressing output indiscriminately. Anthropic’s cost guide covers these context and tooling approaches.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Make requests specific enough to avoid unnecessary exploration
Vague requests can prompt broad scans or work that does not contribute to the desired change. Name the function, file, behavior, or constraint when you know it, and describe the expected result. For example, asking for a change to a named function and the behavior it should have is more bounded than asking Claude to “improve the code.”
Rank #4
For a long or complex task, agree on a plan early enough to catch a wrong direction before it produces more investigation, edits, or output. Specificity should narrow the work, not omit necessary context: include relevant constraints and evidence rather than forcing Claude to guess.
Understand when prompt caching helps
Claude Code automatically uses prompt caching for repeated content such as system prompts. Anthropic’s pricing documentation distinguishes cache writes and reads from ordinary input tokens, so cached tokens do not have the same pricing treatment as uncached input. The benefit depends on how much content repeats and the rates for the model in use; it is not a guaranteed fixed percentage of savings.
Do not restructure a useful workflow solely to chase an assumed caching discount. Compare actual usage for repeated tasks and check the current model-specific rates in Anthropic’s pricing documentation.
Set team controls around the way Claude Code is billed
Teams should first identify how each user accesses Claude Code. A Team or Enterprise plan, Console API usage, and a cloud-provider deployment can have different reporting and spend-control arrangements. Compare the following before choosing a policy:
- Spend visibility: where usage and charges are reported for that access method.
- Limits: which spend or usage controls are available for the relevant plan or deployment.
- Attribution: whether the organization needs usage tied to individual users, projects, or another unit.
For cloud-provider configurations, Anthropic documents OpenTelemetry and gateway options in its cost guidance. Confirm the reporting and controls for the specific deployment rather than assuming the Console API workflow applies.
A practical order for reducing avoidable usage
- Establish a baseline: inspect
/usageand use the billing source appropriate to your account for actual charges. - Stop carrying unrelated history: use
/clearbetween separate tasks, renaming sessions you may need to resume. - Preserve active-task context deliberately: use
/compactwith explicit details to retain, and keep project instructions concise. - Right-size capability: choose the model and reasoning effort for the task’s difficulty, then escalate for work that needs deeper reasoning.
- Remove measured overhead: inspect
/context, disable unused MCP servers, and filter only command output that is not needed. - Compare like with like: review similar tasks after each change, accounting for model, task size, and access method.
This sequence targets repeated or irrelevant tokens first, while keeping the decisions, code facts, and results needed to continue the work.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
Free tools Windows power users keep installed
One-click scans. No signup required.
