Open-source SDLC harness says it cut Claude Code task costs by up to 75%
NeighborhoodOwn8510 · reddit · 2026-07-28
An open-source multi-agent SDLC harness claims up to 75% lower cost than Claude Code
AutoDev Studio turns a plain-English feature request into a staged software lifecycle: a PM agent clarifies scope, a Dev agent codes on an isolated branch, QA runs real tests, and a separate reviewer model checks the diff before a human merges the PR.
Key points:
- On two Python repos (35k and 82k LOC), the tuned pipeline beat a cold single-agent Claude Code run on 6/6 well-localized tasks.
- Reported savings ranged from 7% to 75%, with one example dropping from $6.83 and 207 turns to about $1.70.
- It deliberately uses different model families for planning, coding, and review to avoid self-rubber-stamping.
- The author says the system localizes code using a deterministic symbol map, live grep, and current file contents, so it does not rely on a stale embedding snapshot.
- It can run against Claude Code, Codex, Cursor, Aider, or Gemini CLI using existing logins instead of separate API keys.
The author also notes where it loses: trivial one-line edits can cost more because of the multi-stage overhead, and one hard cross-cutting bug produced a cheaper but narrower fix than the baseline.
Related event: Open-Source AutoDev Studio Claims to Cut Costs by 75%(6 posts)→
More from coding & agent
- Emblem says it can build an 80-slide cited deck in 35 minutes instead of 100 hours — rohanpaul_ai · 2026-07-28
- Claude Opus 5 keeps running Python scripts to edit files in coding workflows — samuelcolvin · 2026-07-28
- Vercel highlights its open-source stack, including Eve and agent-browser — evilrabbit_ · 2026-07-28
- LightOn says its agent search hits 86.27% accuracy with just 9.7 calls — IgorCarron · 2026-07-28
- The Modern Wait Equation: Fix AI Slop Now or Wait for the Next Model? — ilanbigio · 2026-07-28
- Kimi K3 pairs a 2.8T MoE with 1M context and multi-teacher post-training — novasarc01 · 2026-07-28