{"slug": "i-was-tired-of-all-false-claims-so-i-built-atelier-honest-30-saving-on-cc", "title": "I was tired of all false claims, so I built Atelier, honest 30% saving on CC.", "summary": "Atelier, a 30-second install tool that sits underneath Claude Code to reduce token waste, claims 30% savings on AI coding costs by providing better search, shorter file reads, compact command output, and reusable memory. The tool offers a live savings meter and allows users to verify savings by scanning their own local agent history, with benchmark results showing cost reductions of 12.9% to 67% across various tasks.", "body_md": "Atelier is a 30-second install that helps Claude Code waste fewer tokens while you work. Keep using Claude Code normally; Atelier sits underneath it and gives the agent better search, shorter file reads, compact command output, reusable memory, and a live savings meter.\n\n**Live savings across Atelier sessions**\n\n[Install](#install-in-30-seconds) · [Check your savings first](#check-your-own-savings) · [Why trust the numbers?](#why-trust-the-numbers) · [Results](#results) · [Pricing](#pricing)\n\nRun this once:\n\n```\ncurl -fsSL https://install.atelier.ws | bash\n```\n\nThen turn it on inside the project where you use Claude Code:\n\n```\ncd your-project\natelier init\n```\n\nOpen Claude Code like you normally do. Atelier wires in better tools behind the scenes and starts tracking savings as sessions finish.\n\nAlready installed?\n\n```\natelier update\n```\n\nCheck that everything is connected:\n\n```\natelier doctor\n```\n\nAtelier does not ask you to learn a new coding app. It improves the work Claude Code already does:\n\n| Before | With Atelier |\n|---|---|\n| Claude reads broad files and long terminal output. | Claude gets the exact code ranges and compact results it needs. |\n| The same context gets rediscovered again and again. | Useful session context can be reused. |\n| You pay for long explanations inside the working transcript. | Outputs stay shorter while code, commands, filenames, and errors remain exact. |\n| Savings are hard to see. | A local meter shows tokens, cost, and savings adding up. |\n\nThe point is simple: more of your Claude subscription should go into useful work, not repeated setup and paid noise.\n\nDo not take our 30% claim on faith. Before installing, you can scan your own local agent history:\n\n```\ncurl -fsSL https://savings.atelier.ws | bash\n```\n\nWhat it does:\n\n- Reads local Claude/Codex agent session files.\n- Estimates where Atelier would have used fewer tokens or cheaper tool calls.\n- Prints savings from your own history.\n- Uses a temporary local store.\n- Needs no Atelier account and no provider API keys.\n\nUseful variants:\n\n```\ncurl -fsSL https://savings.atelier.ws | bash -s -- --since 30d --top 10\ncurl -fsSL https://savings.atelier.ws | bash -s -- --host codex --limit 20\n```\n\nYou are right to be skeptical. A live badge alone proves very little because anyone can fake a counter.\n\nAtelier uses four checks instead:\n\n**Raw benchmark receipts:** headline numbers link to committed per-task runs, costs, turn counts, setup notes, and reproduction commands.**Your own scan:** the savings command checks your machine, not our marketing page.**Labeled live badges:** live counters show aggregate usage adding up; they are not used as the source of the 30% benchmark claim.**Rows where Atelier does not win:** Terminal-Bench 2.1 is flat on accuracy (-0.2pp) and only cheaper on cost. It stays in the results table.\n\nThe trust is the audit trail, not the animation.\n\nMeasured on the same model, same tasks, and same environment:\n\n| Benchmark | Baseline correct | Atelier correct | Correct delta | Baseline cost | Atelier cost | Cost delta |\n|---|---|---|---|---|---|---|\n| SWE-bench Verified, 50 tasks x 5 reps | 80.8% | 92.8% |\n+12.0 pp |\n$234.84 | $165.45 |\n29.5% cheaper |\n| SWE-bench Lite, 10 tasks x 3 reps | 93.3% | 100% |\n+6.7 pp |\n$12.38 | $10.79 |\n12.9% cheaper |\n| SWE-bench Pro, 10 tasks x 5 reps | 88.0% | 90.0% |\n+2.0 pp |\n$39.01 | $30.61 |\n21.5% cheaper |\n| Exploration tasks across 7 large repos x 5 reps | - | - | - | $19.11 | $6.29 |\n67% cheaper |\n| Telegraphic Q&A, 20 prompts x 5 reps | - | - | - | $8.93 | $5.34 |\n40.2% cheaper |\n| Terminal-Bench 2.1, 89 tasks vs public leaderboard* | 78.9% expected | 78.7% | -0.2 pp | $96.76 | $69.52† |\n28.1% cheaper† |\n\n* Atelier: 1 rep/task. Baseline: public tbench.ai leaderboard, 5-rep average per task. † Other 5 tasks in Atelier timeout and cannot capture cost; see .\n\nSWE-bench Verified detail:\n\n| Metric | Baseline | Atelier | Delta |\n|---|---|---|---|\n| Turns | 6,962 | 4,336 | 37.7% fewer |\n| Output tokens | 3.04M | 2.19M | 27.9% fewer |\n| Wall-clock | 14.3h | 10.9h | 23.7% faster |\n\nRaw runs, setup notes, and reproduction commands live in [BENCHMARKS.md](/atelier-ws/atelier/blob/main/BENCHMARKS.md) and [ benchmarks/codebench/results/](/atelier-ws/atelier/blob/main/benchmarks/codebench/results).\n\nRetrieval quality (MRR) across ~7,200 query/gold pairs on 14 repos -- ripgrep, ast-grep, universal-ctags, Serena, CodeGraph, cocoindex-code, codebase-memory-mcp, fff-mcp, code-index-mcp, and jCodeMunch all scored on the identical corpus:\n\n| Provider | MRR | rec@1 | p95 |\n|---|---|---|---|\nAtelier +semantic (BGE) |\n0.727 |\n0.650 |\n390ms |\n| Atelier lexical (default) | 0.676 | 0.582 | 134ms |\n| cocoindex-code (best rival) | 0.557 | 0.457 | 595ms |\n| serena | 0.401 | 0.359 | 3834ms |\n| ripgrep | 0.376 | 0.320 | 66ms |\n| universal-ctags (worst rival) | 0.237 | 0.226 | 1ms |\n\nNo one had scored these 10 tools against each other on a shared query set before -- each publishes its own number, on its own terms, against its own baseline. Full 13-row table and per-tool \"claim vs. scored\" breakdown: [atelier.ws/vs](https://atelier.ws/vs) · [docs.atelier.ws/benchmarks/results](https://docs.atelier.ws/benchmarks/results).\n\nAlso worth a look, two comparisons outside code search: [rtk](https://atelier.ws/vs/rtk), a 69.6k-star Rust CLI proxy -- self-estimated savings vs. Atelier's accuracy-checked Terminal-Bench number. [\"Just tell Claude to be terse\"](https://atelier.ws/vs/caveman), the free DIY alternative -- benchmarked head-to-head on the same 20 prompts, including the one where it backfires.\n\nClaude is strong, but the work around Claude is often wasteful. Atelier reduces that waste.\n\n**Better inputs:** the agent gets relevant symbols and exact file ranges instead of whole files.**Better outputs:** command output and replies stay compact without losing exact technical facts.**Better memory:** useful context can be reused instead of rediscovered.**Better guardrails:** tools and hooks reduce risky edits, oversized reads, and unverified \"done\" states.\n\nAtelier does not make Claude a different model. It makes the loop around Claude cleaner, which is why the same model solved more tasks in the benchmark.\n\n- Works with Claude Code, Codex, Copilot, Cursor, opencode, Hermes Agent, LangChain, the OpenAI SDK, Gemini ADK, and any other MCP-compatible coding agent.\n- Runs locally by default.\n- Apache-2.0 open source runtime.\n- No account needed to start.\n- Live local stats for cost, tokens, and savings.\n- Optional paid features for heavy users and teams.\n\n[Installation](/atelier-ws/atelier/blob/main/docs/installation.md)[Troubleshooting](/atelier-ws/atelier/blob/main/docs/troubleshooting.md)[Benchmarks](/atelier-ws/atelier/blob/main/BENCHMARKS.md)·[full results, backed by docs](/atelier-ws/atelier/blob/main/docs/benchmarks/results.md)·[every \"vs\" comparison, with sources](https://atelier.ws/vs)[CLI reference](/atelier-ws/atelier/blob/main/docs/cli.md)[Architecture](/atelier-ws/atelier/blob/main/docs/architecture.md)\n\nApache 2.0 — see [ LICENSE](/atelier-ws/atelier/blob/main/LICENSE).", "url": "https://wpnews.pro/news/i-was-tired-of-all-false-claims-so-i-built-atelier-honest-30-saving-on-cc", "canonical_source": "https://github.com/atelier-ws/atelier", "published_at": "2026-07-09 02:22:31+00:00", "updated_at": "2026-07-09 02:42:11.603000+00:00", "lang": "en", "topics": ["ai-tools", "developer-tools", "large-language-models", "ai-agents"], "entities": ["Atelier", "Claude Code", "Anthropic", "SWE-bench", "Terminal-Bench"], "alternates": {"html": "https://wpnews.pro/news/i-was-tired-of-all-false-claims-so-i-built-atelier-honest-30-saving-on-cc", "markdown": "https://wpnews.pro/news/i-was-tired-of-all-false-claims-so-i-built-atelier-honest-30-saving-on-cc.md", "text": "https://wpnews.pro/news/i-was-tired-of-all-false-claims-so-i-built-atelier-honest-30-saving-on-cc.txt", "jsonld": "https://wpnews.pro/news/i-was-tired-of-all-false-claims-so-i-built-atelier-honest-30-saving-on-cc.jsonld"}}