← back
>_ CLAUDE-AHOLICS · FREE
SAVE A FORTUNE
ON CLAUDE
Every legit way to spend less and get more. No snake-oil.
PLAN = makes your Pro/Max plan last longer $$$ = cuts your API / Claude Code bill
The one thing to understand
Every message re-bills the entire conversation history as input. Long chats get expensive fast — that's why "start a new chat" is the #1 tip. And output tokens cost ~5× input, so short answers save the most.
Quick wins
Start a new chat per topic. Long threads re-bill everything, every turn.$$$PLAN
Ask for the length you want. "3 bullets." "One line." Output is the pricey part.
Use the right model. Haiku for simple tasks, Sonnet for most, Opus only when needed.$$$
Don't make it echo your input. Ask for just the change, not a full re-print.
Turn off extended thinking for simple tasks. Thinking = billable tokens.
Summarise & restart. When a chat balloons: "summarise the key points" → new chat → continue.
Keep standing context lean. Project files & system prompts load every message. Trim them.
The HUGE ones (API / Claude Code)
Prompt caching ★ Cache a reused prefix (docs, system prompt) — cached reads cost ~10% (≈90% off). The single biggest lever.$$$
Batch API — 50% off. Bulk, non-urgent jobs at half price.$$$
Model routing. Draft cheap, escalate only when needed.$$$
RAG, don't dump. Send only the relevant chunks, not the whole manual.$$$
Cap max_tokens & trim tool results. Stop runaway output; don't feed back giant JSON.$$$
Claude Code & Pro/Max
/clear & /compact between tasks, to drop context you're re-billing.
Point at files/line-ranges, don't dump whole repos.
New chats early (Pro/Max) = the #1 way to not hit your limit.PLAN
Downscale images before sending. Cost scales with resolution — crop & shrink.
The Token-Saving Bible (Pro)
Every tactic explained in full — with worked examples, diagrams and copy-paste setups (prompt-caching template, batch, model-router). A proper guide, not a listicle. Cut your bill, keep every capability.
Get the Pro version → £12