Launchfiles logoLaunchfiles

Planned — not available yet. Nothing on this page runs today.

Code review automation · private beta

A review gate that blocks bad AI agent PRs before they hit main.

A fail-closed review gate for AI agent pull requests. Checks file scope, tests, types, and security. Signed receipt. You still approve the merge.

Private beta · No card required · No auto-merge

Scoped file boundary
Verifiable test receipt
Human approval boundary
Next safe action

Contrast

Same agent. Different outcome.

AI agents write code fast, but without deterministic boundaries they drift, relax assertions, and delete failing tests to force a green check.

Without a review gate

  • Unclear prompts
  • Hallucinations
  • Accidental rollbacks

With a review gate

  • Clear boundaries
  • Verified receipts
  • Human approval

Deterministic checks

Four gates that block bad agent PRs

Launchfiles executes deterministic verification checks before any human review. No vibe checks. No soft green.

01

Diff Scope Containment

Verifies git diff stays strictly within declared file boundaries. Rejects unapproved changes to configs, lockfiles, or random utils.

PASS (2 files modified: auth.ts, session.ts · 0 out-of-scope files)
02

Test Assertion Preservation

Executes test runner in an isolated sandbox. Blocks any PR where existing assertions were deleted, commented out, or weakened to force a pass.

PASS (12/12 tests passed · 0 tests deleted · 0 assertions relaxed)
03

Strict Typecheck Compilation

Runs strict TypeScript compilation (tsc --noEmit). Blocks silenced compiler errors, loose type casts, or newly introduced any types.

PASS (tsc --noEmit clean · 0 compiler errors · 0 widened any types)
04

Security & Credential Scan

Static analysis and SARIF scan verify zero leaked tokens, unpinned dependencies, ambient system calls, or unauthorized privilege elevation.

PASS (SARIF clean · 0 hardcoded secrets · 0 ambient authority elevations)

Inspect the signed execution receipt.

Every gate run produces a receipt with full check logs, diff hashes, and the human approval requirement.

  • Private beta · No card required · Never auto-merges
  • Compatible with Cursor, Claude Code, GitHub Copilot, and CLI agents.
Fail-closed quality gate · Human approval boundary
Start free →

Workflow integration

From agent diff to safe merge

  1. 01

    Bound the mission

    Name one decision, the evidence available, the owner, and actions that remain forbidden.

  2. 02

    Return the artifact

    Produce the route-specific brief, work orders, gates, and honest evidence gaps—not a transcript.

  3. 03

    Expose the boundary

    Show what ran, what did not run, what needs approval, and the next safe action.

  4. 04

    Keep design claims honest

    Page repair, design-system checks, and landing briefs need DOM and screenshot evidence before any design-compliant claim.

Honest fit

When this helps (and when it is not a fit)

When this helps
Teams where AI agents generate pull requests that need automated verification before human review.

Launchfiles is designed for technical founders and engineering leads who want verifiable proof, not AI optimism.

Structured answers

Frequently asked questions

Does Launchfiles automatically merge approved PRs?
Never auto-merges. Launchfiles follows a strict fail-closed, production-no-publish doctrine. It presents pre-verified change packages with passing gate receipts, leaving the final merge authorization to the engineering lead or founder.
What is an AI PR review gate?
An AI PR review gate is a deterministic verification harness that evaluates agent-generated pull requests against strict architectural rules, test passes, and security policies before code reaches staging or production.
Is this a project-management stage-gate or phase-gate review?
No. This review gate is for AI agent pull requests: fail-closed checks on the diff, tests, types, and secrets before a human approves merge. It is not a stage-gate or phase-gate process for projects.
How does an automated PR review gate differ from standard CI/CD?
While CI/CD validates build syntax and general tests, an AI PR review gate specifically enforces work order file scope boundaries, detects unapproved dependency injections, and verifies that AI code adheres to repository doctrine.

Fail-closed quality control

Ship agent code with proof, not hope.

A fail-closed review gate for AI agent pull requests. Checks file scope, tests, types, and security. Signed receipt. You still approve the merge.

Private beta · No card required · Never auto-merges