If Claude Code seems to be using too many tokens, first find out what you are looking at: session usage, subscription usage, an API charge, or a cloud-provider bill. In Claude Code, run /usage to inspect usage and /context to see what is occupying context. Then check conversation history, project instructions, MCP tools, and the selected model before changing plans or buying anything.
First, tell usage displays apart
Claude Code’s /usage command does not mean the same thing for every billing route. Its display is useful for diagnosing activity, but it is not always an invoice.
![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)
As an Amazon Associate I earn from qualifying purchases.
| What you use | What /usage shows |
Where to verify billing |
|---|---|---|
| API access | The Session block reports detailed token usage for the current session. The displayed dollar amount is calculated locally and is an estimate. | Claude Console’s Usage page is the authoritative place to check Anthropic API billing. Anthropic’s cost guide |
| Claude Pro or Max | Plan usage bars, activity information, and a usage breakdown. A session cost figure is not a subscription invoice. | Check usage in the relevant subscription account; do not interpret the local session cost as an additional subscription bill. Anthropic’s cost guide |
| Bedrock, Google Cloud Agent Platform, or Microsoft Foundry | Claude Code can be routed through these providers, so Anthropic Console may not be where the charge appears. | Check the billing account for the cloud provider handling the request. Anthropic’s LLM gateway guide |
| Gateway or shared credentials | The usage may accrue to the owner of the configured credentials rather than directly to the person using the CLI. | Ask the credential or gateway owner which account receives the charge. Anthropic’s LLM gateway guide |
If you are unsure which route is active, check how Claude Code was configured and whose credentials it uses. A plan usage bar, a local token estimate, and a provider’s billed spend are different measurements.
Use /context to find what is growing
Run /context in Claude Code. It helps show what is occupying the current context, making it a better starting point than guessing from a usage bar alone. Anthropic puts the key relationship plainly: “Token costs scale with context size: the more context Claude processes, the more tokens you use.” Anthropic, “Manage costs effectively”
#1 Best Overall
Look for a long conversation, repeated broad instructions, large tool results, or project guidance that the task does not need. Usage can rise because Claude is processing more context, even if your latest message is short.
Check the conversation before starting another task
Previous conversation history remains part of the working context. If you are switching to unrelated work, use /clear to start without carrying that history forward. If you are continuing a task where earlier decisions matter, use /compact and give Claude instructions about what to preserve. Clearing removes an unrelated thread; compacting is for keeping a useful thread in a smaller form. Anthropic’s cost guide
Inspect project instructions, not just chat history
Claude Code can read CLAUDE.md and CLAUDE.local.md instruction files from the working directory and parent directories. Instructions in child directories may also be loaded when Claude reads files there; instruction files can import other files as well. Long or duplicated guidance can add to the context used for a task. Anthropic’s memory guide
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Scan for outdated or missing drivers - takes under a minute3Clear out junk files and repair common Windows errors- Inspect the instruction files that apply to the current directory and task.
- Check for duplicated guidance or imported files that repeat material.
- Do not delete instructions simply because they are long: project-specific rules may be essential for safe or correct work.
Review MCP servers and tool output
MCP connects Claude Code to external tools and data. Tool descriptions and results can contribute to context, and a high-output tool may return far more information than a task requires. Use /mcp to review configured servers, then disable or avoid servers that are not needed for the current work. Anthropic’s MCP documentation discusses high-output tools and a configurable maximum; check the current documentation and your installed version before changing related settings. Anthropic’s MCP guide
Confirm which model is selected
Run /model to see or change the model in the CLI. You can also select a model when starting Claude Code with --model. Anthropic recommends Sonnet for most coding tasks and reserving Opus for complex architectural or multistep reasoning. Model availability and relative economics can change, so check the current selection rather than relying on an old price comparison. Anthropic’s cost guide CLI reference
A quick troubleshooting sequence
- Run
/usageand identify whether you see API session details or subscription usage information. - Run
/contextand note the largest contributors to the current context. - If the next task is unrelated, use
/clear; if you are continuing, use/compactwith explicit preservation instructions. - Review applicable
CLAUDE.mdandCLAUDE.local.mdfiles and their imports for unnecessary duplication. - Use
/mcpto check whether configured servers are needed and whether tool results are unusually large. - Use
/modelto confirm the active model. - Verify charges with the account that actually bills the request: Claude Console for Anthropic API usage, the subscription account for plan usage, or the relevant cloud provider or gateway owner for routed usage.
These checks address context and billing visibility. The available evidence does not establish a particular computer, accessory, or system-cleanup product as a way to reduce Claude Code token use.
Quick Recap
Best Value
- Book - 1, 000 books to read before you die: a life-changing list (1000 before you die)
- Language: english
- Binding: hardcover
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.
