ClaudeHack

#cost

6 hacks

  1. 1.
    Tune /effort so depth matches the task, not every turn(claude.com)

    Use /effort (low→max) or CLAUDE_CODE_EFFORT_LEVEL so simple edits stay cheap and hard debugging gets deeper reasoning without paying max on every prompt.

    0 points·13 hours ago·Terminal / Productivity / Cost Optimization·easy·discuss
  2. 2.
    Use /usage to find which MCP tools and skills burn budget(code.claude.com)

    Inspect `/usage` (alias `/cost`) for per-skill, subagent, and MCP breakdowns, then cut the expensive outliers.

    0 points·13 hours ago·Usage & Limits·easy·discuss
  3. 3.
    Enter plan mode before large refactors to avoid rework(code.claude.com)

    Cycle into plan mode (Shift+Tab) so Claude explores and proposes an approach for approval before editing.

    0 points·13 hours ago·Terminal / Productivity / Cost Optimization·easy·discuss
  4. 4.
    Move PR and migration workflows out of CLAUDE.md into skills(code.claude.com)

    Shift specialized playbooks into on-demand skills so every session does not pay for instructions you rarely need.

    0 points·13 hours ago·Terminal / Productivity / Cost Optimization·medium·discuss
  5. 5.
    Disable unused MCP servers to reclaim startup tokens(code.claude.com)

    Turn off MCP servers unrelated to the current project—their tool schemas still consume context at session start.

    0 points·13 hours ago·MCP / Agents·easy·discuss
  6. 6.
    Enable automatic prompt caching on multi-turn API chats(docs.anthropic.com)

    Add a top-level cache_control ephemeral flag so growing conversation prefixes reuse cached tokens and cut API cost and latency.

    0 points·13 hours ago·Usage & Limits·intermediate·discuss