Season 3 · CLI Launch & Loop Engineering · Jun 30-Jul 7

Build the Loop.

Your agent writes the code. The open-source TestSprite CLI checks it: real tests against your live app, with verdicts your agent acts on. Write, verify, fix, verify.

View Past Projects
TestSprite CLI verifying an app from the terminal

$5,000

Prize Pool

7

Days to Build

5

Winners

Open Source

CLI · Apache 2.0

// Prize Pool

$5,000 total.

Build the best loop, or make the CLI better while you build.

Two ways to win

Project Awards

$3,000 · 5 winners

Best projects built with the CLI in a real testing loop. Judged on the loop, not polish or pitch.

Prize tiers: 1st place $1,500 Grand Champion, 2nd $750, 3rd $450, 4th and 5th $150 Honorable Mention

CLI Improvement Bonus

$2,000 pool · standing bounty

Separate from judging. Cash for merged improvements, doesn't affect your Project Award score.

CLI improvement bonus: $100+ per merged PR, more for substantial ones, paid first-come from the $2,000 pool until exhausted
// The Loop

Four Steps.
One Repeats.

WriteMaker

Your coding agent ships code. Claude Code, Codex, Antigravity, whatever you build with.

VerifyChecker

The TestSprite CLI runs real tests against your live app and returns verdicts.

FixMaker

Agent reads the failure bundle and fixes the root cause.

Verify AgainChecker

Rerun. Pass banks. Then back to the top.

Makeryour coding agentCheckerTestSprite CLI

a loop with no real checker doesn't fail loudly. It hallucinates progress.

// Start in 3 minutes

Install. Onboard.
Loop.

One install, one init, and the CLI installs a verification skill into your agent, so it knows when and how to run the checker. Full commands & docs live in the open-source repo.

01 // Install

Get the CLI

View on GitHub →

02 // Onboard

testsprite init

API key + agent skill. Use your hackathon promo code.

03 // Target

Your live URL

The CLI tests in the cloud, point it at your deployed app.

04 // Loop

create · fix · rerun

The agent drives the loop. Every pass banks.

// How to Submit

What Goes Where.

The agent writes LOOP.md.

One plain-English line per iteration: maker first, then what ran, what broke, what got fixed. Your agent writes it as the loop runs. No hand-writing. Judges read this first.

Backed by your commit history + the platform run history, so the log is fast to read and easy to trust.

Agent-written LOOP.md: one line per run with maker, what ran, and pass/fail verdict

Two steps to submit

Your repo carries the proof the loop ran. Your Discord post is the entry that registers you. Both need to be in before the deadline.

01 // In your GitHub repo

Your source, an agent-written LOOP.md, and a README with the app and live URL.

Project structure: src, .testsprite/failure bundles, agent-written LOOP.md, README.md with app and live URL

02 // Post in Discord#hackathon-submissions

Your live URL, public repo link, and TestSprite account. Demo video optional.

Submission checklist: live public URL, public repo link, TestSprite email and name, demo video link
// Prerequisites

The Checklist.

TestSprite CLI

Required

Install from GitHub · Node ≥ 20 · paid plan via promo code.

Public URL

Required

The CLI tests in the cloud, no localhost. Deploy early, keep it live the whole build.

Public Repo

Required

Source in a public repo. Commit history is the proof the loop ran.

LOOP.md

Required

Agent-written, one line per iteration. No log = invalid submission.

README

Required

App, live URL, and what the loop covered.

Demo Video

Optional

Encouraged. Link it in Discord. Boosts ranking.

// Judging Criteria

Human Judges.

40 pts

Project Quality

Craft, polish, completeness. Does the live app actually work well?

40 pts

Loop Quality

Did a real loop run, and catch and fix real things? Read from LOOP.md, commits, and runs.

20 pts+5

Innovation

Creativity of the project or the loop design. Plus 5 for wiring the checker into CI/CD.

Bonus

Engagement

Discord polls, X shares, and long-form write-ups about your loop.

// Go further

Wire the checker
into CI/CD.

+5 Innovation points · optional

The loop, on autopilot.

Gate a GitHub Actions pipeline on TestSprite. Every push reruns your tests and fails the build if something breaks. One workflow file, one secret. The stickiest way to keep the checker working long after the hackathon.

Read the CI guide →
# .github/workflows/testsprite.yml
on: pull_request
env:
  TESTSPRITE_API_KEY: ${{ secrets.TESTSPRITE_API_KEY }}
  PROJECT_ID: proj_xxxxxxxx
steps:
  - run: npm i -g @testsprite/testsprite-cli
  - run: testsprite test run --all --project "$PROJECT_ID" \
           --wait --output json
    # non-zero exit fails the build ↓
  # --all = backend tests; for frontend, pass a test-id
// Execution Timeline

One Week to Ship.

// Season 03
  1. Jun 30 · 5:00 PM PDT

    Launch & Setup

    Install the CLI, onboard your agent, deploy your app.

  2. Jun 30 – Jul 7

    Build & Loop

    Run write → verify → fix against your live app.

  3. Jul 7 · 4:59 PM PDT

    Submissions Close

    Live URL + repo + LOOP.md + account.

  4. Jul 8 – Jul 12

    Review

    Every entry against the rubric.

  5. Jul 13

    Winners

    Top 5 revealed on X & Discord.

  6. No end date

    CLI Improvement Bonus, always open

    Separate from the hackathon clock. The $2,000 improvement bounty runs until the pool is exhausted. Open a PR improving the CLI any time, even after Jul 13.

    ongoing
// Join

Two steps
and you're in.

Step 1// Follow @TestSprite on X

Announcements, countdowns, and winner reveals go out on X first.

Follow on X

Step 2// Join the Discord

Rules, submissions, polls, Q&A. CLI quickstart pinned in #hackathon-info.

Join Discord

Build the Whole Loop.

Including the half most people skip. Install the CLI, wire it in as your checker, then let your agent loop. Follow on X and join Discord, where everything happens.