What developers are saying about Claude bills.

Live-curated Hacker News threads on Claude costs, Anthropic pricing, and cost reduction — auto-refreshed every few hours. See what real engineers are dealing with, right now.

Last updated: 2026-04-21T22:00:44.578Z · 20 threads tracked

See your own Claude savings → Free audit at aiusage.ai
TL;DR across the threads below: Developers are reporting monthly Claude bills between $100 and $2,500+ and asking how to reduce them. Most workloads have 70–85% waste (retries, redundant context, agent loops). See the 6 verified savings cases →

Recent HN discussions

Tell HN: Anthropic no longer allowing Claude Code subscriptions to use OpenClaw
1099 points · 829 comments · by firloop · 2026-04-03 · view on HN →
I've been waiting over a month for Anthropic to respond to my billing issue
433 points · 200 comments · by nickvec · 2026-04-08 · view on HN →
Show HN: cmux - Ghostty-based terminal with vertical tabs and notifications
198 points · 77 comments · by lawrencechen · 2026-02-19 · view on HN →
Claude Code bug can silently 10-20x API costs
66 points · 7 comments · by wg0 · 2026-03-31 · view on HN →
Show HN: Daemons – we pivoted from building agents to cleaning up after them
44 points · 27 comments · by rileyt · 2026-04-21 · view on HN →
Tell HN: I cut Claude API costs from $70/month to pennies
40 points · 30 comments · by ok_orco · 2026-01-26 · view on HN →
Show HN: Trust Protocols for Anthropic/OpenAI/Gemini
40 points · 33 comments · by alexgarden · 2026-02-18 · view on HN →
Show HN: SuperDesign.Dev – first Open-source Design Agent lives inside IDE
24 points · 8 comments · by jzdesign1993 · 2025-06-25 · view on HN →
Anthropic bans orgs without warning
17 points · 5 comments · by alpinisme · 2026-04-21 · view on HN →
Show HN: Navigator Mode (Like Claude Code) for Aider
16 points · 5 comments · by tekacs · 2025-04-13 · view on HN →
Show HN: Page Magic: Use AI to customize any web page
15 points · comments · by khaledh · 2025-06-03 · view on HN →
Show HN: CivBench a long-horizon AI benchmark for multi-agent games
12 points · 24 comments · by mbh159 · 2026-02-25 · view on HN →
Show HN: I built a 55K-word email marketing knowledge base and Claude Code skill
10 points · 2 comments · by cosmoblk · 2026-02-21 · view on HN →
Show HN: OpenClawdex – Open-Source Orchestrator UI for Claude Code and Codex
7 points · 7 comments · by alekseyrozh · 2026-04-19 · view on HN →
Anthropic's pricing wall is routing enterprise revenue to OpenAI
6 points · 2 comments · by luckygreen · 2026-02-18 · view on HN →
Show HN: I cut my Claude API bill by 66% with Git-based context
5 points · comments · by alessandro-a · 2025-08-28 · view on HN →
Show HN: Zyg – Stop Writing Status Updates
5 points · 4 comments · by flyingsky · 2025-09-02 · view on HN →
Ask HN: Isn't AnthropicAI Claude 3's price too expensive?
5 points · 1 comments · by behnamoh · 2024-03-09 · view on HN →
Built my own newsletter platform, from 0 to 5k subs in 10 days
5 points · 3 comments · by loverajoel · 2024-03-14 · view on HN →
Show HN: Pencil Puzzle Bench – LLM Benchmark for Multi-Step Verifiable Reasoning
5 points · comments · by bluecoconut · 2026-03-03 · view on HN →

Pattern across every thread

One shape keeps repeating. Dev opens Anthropic dashboard. Bill is 3–10× what they expected. Debug reveals most of it is retries on malformed output, agent chains re-reading the same context, and support automation burning tokens on the same "have you tried restarting it" style replies.

The fix isn't switching providers. The fix is auditing where the bill actually goes — and then cutting the waste on the same prompts. Six verified cases so far:

Run the free audit on your own usage →