bolt Fully Automated · $0 Cost

TestFlight Feedback
Auto-Fix Pipeline

Testers submit feedback in TestFlight. AI reads it, fixes the code, verifies the build, and ships a new version — all automatically, every 15 minutes.

check_circle Claude Code (any plan) check_circle Self-hosted check_circle MIT License
feedback-autofix.yml

Fetch feedback — Found 3 new items

Claude auto-fix — 3 commits

fix: card height inconsistency when value is zero

fix: remove redundant calorie display area

fix: default collapse assumptions section

Build verification — BUILD SUCCEEDED

Unit tests — 268 tests passed

Push and deploy — TestFlight build triggered

Done in 4m 12s — $0 cost

account_tree

How It Works

End-to-end automation from tester feedback to new TestFlight build. No human intervention required.

feedback

Step 01

Tester Feedback

Screenshots & crash reports from TestFlight

download

Step 02

Fetch & Filter

Poll App Store Connect API every 15 min

smart_toy

Step 03

AI Fix

Claude Code reads codebase & commits fixes

verified_user

Step 04

Build & Test

xcodebuild + unit tests. Reverts if fail

flight_takeoff

Deploy

TestFlight

Auto-push to release branch → new build

Quick Setup

One command to deploy with Claude Code or Codex

terminal

Claude Code

claude -p "Install testflight-feedback-fix from github.com/nicejungle/testflight-feedback-fix into this project"
code

Codex

codex "Install testflight-feedback-fix from github.com/nicejungle/testflight-feedback-fix into this project"

Manual Setup

1 Install Runner
git clone github.com/nicejungle/
testflight-feedback-fix
cd testflight-feedback-fix
./scripts/setup-runner.sh
2 Copy to Project
cp workflows/* .github/workflows/
cp -r lib/ scripts/ .
cp fastlane/Fastfile fastlane/
3 Add Secrets
gh secret set ASC_KEY_ID
gh secret set ASC_ISSUER_ID
gh secret set ASC_KEY_CONTENT
gh secret set ASC_APP_ID
gh secret set KEYCHAIN_PASSWORD
attach_money Cost: $0
timer Poll: 15 min
build Build + Test Gate
verified MIT Licensed

Key Features

  • check_circleAI code fixes via Claude Code (Opus 4.6)
  • check_circleBuild + test with 3x auto-retry
  • check_circleReal-time dashboard via Tailscale
  • check_circleLive Claude thinking stream
  • check_circleClick to view commit diffs & run logs
  • check_circleDismiss, fix single, trigger, cancel
  • check_circleAdaptive fetch (20→200 items)
  • check_circle$0 cost — self-hosted, any plan
translate

简体中文

全自动化流水线:TestFlight 反馈 → AI 修复 → 构建验证 → 自动发版。使用 Claude Code 订阅额度,零额外费用。

git clone github.com/nicejungle/
testflight-feedback-fix
./scripts/setup-runner.sh
完整文档 open_in_new
translate

日本語

完全自動パイプライン:TestFlightフィードバック → AI修正 → ビルド検証 → 自動デプロイ。Claude Codeサブスクリプションを使用、追加費用ゼロ。

git clone github.com/nicejungle/
testflight-feedback-fix
./scripts/setup-runner.sh
ドキュメント open_in_new

FAQ

How much does this cost? expand_more
$0 extra. Claude Code runs on your existing subscription (any paid plan). The Mac Mini is yours. GitHub Actions is free for self-hosted runners.
What if Claude introduces a bug? expand_more
The pipeline runs xcodebuild + unit tests after every fix. If either fails, all changes are automatically reverted. Nothing broken ever gets pushed.
What if there's no new feedback? expand_more
The workflow checks, finds 0 new items, and exits in ~3 seconds. Claude is never called. No resources consumed.
Can I use this without Claude Code? expand_more
Yes — remove the "Claude auto-fix" step and use it as a manual feedback-to-deploy pipeline. Or swap in any AI tool that has a CLI.