Recommended Free Tools
In one author-reported test using Claude Code 2.1.285 in an empty directory, the first request contained 23,207 input tokens by default and 2,664 after adding --strict-mcp-config --tools "". Those are measurements from one configured machine and account, not a fixed Claude Code overhead or an Anthropic-published specification. The test’s total includes uncached input, cache-read input and cache-creation input.
What counts as Claude Code’s first-request tokens?
The figures come from Rulestack’s October 3, 2026 report of a headless Claude Code test. The author calculated the first-request total as input_tokens + cache_read_input_tokens + cache_creation_input_tokens in the response’s usage block. It therefore counts input whether newly processed or served from cache; it is not simply the size of the user’s prompt.
As an Amazon Associate I earn from qualifying purchases.
The test ran claude -p "Reply with just OK." with stream JSON output, verbose mode, a one-turn limit, Opus selection and a settings override disabling hooks. The author says ten configurations were each run twice on September 30, 2026; each pair matched to the token and each run returned OK. These are the author’s reported experimental details, not an independent audit. Read the report and its methodology.
Free tools Windows power users keep installed
One-click scans. No signup required.
Why does an empty folder still produce a large request?
An empty working directory does not mean an empty Claude Code session. The tested account and user setup supplied configuration and context beyond project files. In the default run, the author reports that the session knew about 78 tools, 48 skills, five MCP servers and five agent types. Eleven built-in tools were sent with full definitions; other tool names or descriptions and attached context also contributed.
#1 Best Overall
The author identifies context for the skill listing, deferred tool names, MCP instructions, agent descriptions, account and session, permission mode, environment, model, date and a token-budget reminder. The setup also had a plugin, an MCP server, skills, connected account connectors and auto permission mode. These counts and components describe that particular setup; they should not be assumed for another user’s account or machine.
What changed in the author’s step-by-step test?
The report shows a sequence of changes from the same base run. Each row’s reduction is relative to the preceding row, not a standalone price for the named flag or setting.
| Configuration in the reported sequence | First-request total | Reduction from prior step |
|---|---|---|
| Default | 23,207 | Baseline |
ENABLE_CLAUDEAI_MCP_SERVERS=false |
20,707 | 2,500 |
Also add --strict-mcp-config |
20,316 | 391 |
Also disallow Skill and Workflow |
11,161 | 9,155 |
Instead use --tools "Bash,Read,Edit,Write" with strict MCP configuration |
5,671 | 5,490 |
Use --strict-mcp-config --tools "" |
2,664 | 3,007 |
The 2,664-token result is the final comparison named in the headline: it uses the two flags together. The intermediate steps show that the reported total changed as more configuration was removed or restricted. Because prompt components can interact, the same sequence in another order could assign different reductions to individual changes; these row-by-row differences should not be treated as general-purpose savings for a flag.
What remains in the 2,664-token request?
Removing the reported tools and MCP context did not reduce the first request to the user’s short instruction. The report says the remaining context still included the system prompt, environment and model/date notes, an account note and the user prompt. The measurement is the total input recorded for the first request, not a count of visible project files or words typed by the user.
Rank #3
When might restricting tools help?
For a scripted job that needs only shell and file access, the test suggests that explicitly restricting built-in tools can reduce first-request context in a setup like the author’s. The measured 5,671-token row used --tools "Bash,Read,Edit,Write" with strict MCP configuration; the 2,664-token row used an empty tool list alongside strict MCP configuration. These are results from that test, not a guarantee for current or different installations.
- If a job needs skills, it must retain the Skill tool; the author’s experiment associated the skill listing with additional context.
- If a job should not inherit account connectors, the report suggests considering strict MCP configuration. A job that needs specific MCP servers may require an explicit MCP configuration.
- The author also tried
--setting-sources project,localand--safe-mode; in this setup, the author interpreted their effects as removing parts of user customizations. That observation does not establish a general behavior across machines or accounts.
Check the current CLI reference and your own configuration before adopting a command-line combination; the report’s quoted CLI behavior was not separately verified against current official documentation.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.How broadly do these numbers apply?
They are a case study, not a benchmark for every Claude Code user. They reflect one model selection, Claude Code 2.1.285, one machine and account configuration, and the author’s particular loaded skills and connectors. Tokenization and available context vary with those inputs.
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 problemsThe author says transcript internals used to inspect prompt attachments and snapshots are undocumented and may change. The report also notes that the paired runs do not rule out connector variability. It did not test --bare because the machine used subscription login rather than an API key, and did not test several other flags and settings. An earlier experiment differed in version and other conditions, so its token difference cannot be attributed solely to a release change.
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.
