Skip to main content
CodePeelCodePeelDocumentation
Back

Get Started

Getting StartedFeaturesTeamsBilling

Integrations

VS Code ExtensionMCP Server@codepeel Chat

Configuration

.codepeel.ymlPre-merge ChecksLearnings

Automation

Auto-TestAuto-Fix

Reference

SecurityFAQ

Frequently Asked Questions

Everything you need to know about CodePeel's AI code review engine, data security, model infrastructure, pricing, and integrations.

1. Accuracy, Quality & Noise Reduction

How does CodePeel prevent false positives and review spam?

CodePeel is engineered with a precision-over-recall architecture to eliminate reviewer fatigue. Rather than flooding your PR with hundreds of subjective formatting nits, CodePeel runs multiple quality gates:

  1. Slop & Fluff Detection: Heuristic and LLM filters strip out generic commentary ("Consider refactoring this", "Looks good overall") and keep only actionable findings.
  2. Parallel Validation: Every finding is validated against static analysis (SAST) and security heuristic scanners before being posted.
  3. Per-File & Per-PR Density Caps: Prevents comment storms by capping findings per file hunk.
  4. Interactive Learning: If CodePeel flags something you disagree with, reply with @codepeel ignore: <reason> to permanently suppress that pattern across your repository.

What review profiles are available?

You can tune the aggressiveness of CodePeel via .codepeel.yml or the web dashboard:

  • chill: Only flags critical security vulnerabilities, severe logic bugs, and high-impact memory/performance leaks. Zero style or architectural opinions.
  • balanced (Default): Catches security vulnerabilities, bugs, edge-case regressions, and high-confidence anti-patterns.
  • strict: Exhaustive analysis including test coverage gaps, edge-case validation, and architecture convention checks.
# .codepeel.yml
auto_review:
  profile: chill # Options: chill | balanced | strict

Can CodePeel learn my team's unique coding conventions?

Yes. CodePeel features an active repository learning engine:

  • Teach preferences: @codepeel learn: Always use Zod schemas for external API validation
  • Suppress noise: @codepeel ignore: Ignore missing return type annotations in test files

Learnings are stored per-repository in your Knowledge Base and automatically injected into subsequent review context windows.

2. Data Privacy & Security

Does CodePeel store or train on my proprietary code?

No. CodePeel enforces a strict Zero Data Retention policy for source code:

  • Ephemeral Processing: Code diffs are pulled in-memory exclusively to perform the review and are completely discarded after the analysis cycle finishes.
  • Zero Model Training: CodePeel uses enterprise zero-data-retention agreements with LLM providers. Your code is never used to train or fine-tune public or private models.
  • Metadata Only: Only high-level review metadata (finding titles, severity counts, line numbers, and health scores) is stored in your encrypted dashboard database.

What permissions does the GitHub App require?

CodePeel requests the minimum permissions required to review pull requests and apply fixes:

PermissionAccessWhy it's needed
Pull RequestsRead & WriteTo read diffs, post inline review comments, and create auto-fix/test PRs.
ContentsRead OnlyTo read .codepeel.yml configuration and repository file structure.
Commit StatusesRead & WriteTo post pre-merge quality gate statuses (codepeel/premerge).
Issues / DiscussionsRead & WriteTo listen for @codepeel mentions and reply to developer questions.

CodePeel never requests administrative access, repository deletion, or write access to your repository code branches (except when creating auto-fix branches you explicitly request).

3. AI Architecture & Infrastructure

How does CodePeel's AI engine analyze code?

CodePeel uses a proprietary Multi-Model Hybrid Architecture optimized specifically for automated code review:

  • Fast-Pass Syntax & Security Layer: Ultra-low latency models perform immediate syntax validation, secret scanning, and security vulnerability detection in seconds.
  • Deep-Context Architecture Engine: Large-context models analyze whole-PR dependency trees, cross-file imports, and complex system-level architectural impact.
  • High-Precision Logic Verification: Specialized code-reasoning models verify race conditions, memory safety, logic regressions, and edge cases before outputting findings.

Do I need to provide my own API keys?

No. CodePeel is fully managed. All AI compute, token budgets, and automatic model failovers are included with your subscription. You do not need to manage API keys, quotas, or provider rate limits.

What happens if an AI provider experiences downtime?

CodePeel's routing infrastructure maintains real-time health checks, redundancy pools, and rate-limit cooldown management. If any underlying compute cluster experiences latency or downtime, requests automatically fail over to backup clusters within milliseconds without delaying your pull request checks.

4. Pricing, Quotas & Fair Billing

How does CodePeel pricing work?

CodePeel offers simple, transparent tiers without per-seat lock-in:

  • Free ($0/mo): 30 reviews/month on unlimited public and private repositories, full inline reviews, secret scanning, VS Code extension, and MCP integration.
  • Pro ($9.99/mo or $99.99/yr): 500 reviews/month, automated fix PRs (auto_fix), automated unit test generation (auto_test), custom regex rules, and repository learnings.
  • Max ($24.99/mo or $249.99/yr): Unlimited reviews (subject to a 5,000/mo fair-use cap), priority multi-model routing, and high-concurrency limits.

Do you charge per repository or per seat?

We charge per active subscriber, not per repository. You can connect as many public and private repositories as you want on every plan tier.

What counts as a review?

  • Opening or updating a GitHub Pull Request consumes 1 review.
  • Running a review via the VS Code extension or MCP server (review_code) consumes 1 review.
  • @codepeel chat replies, shortcut commands (learn:, ignore:), and credit balance queries (check_credits) are free and do not consume review quota.

5. Setup, Permissions & Troubleshooting

How do I install CodePeel on GitHub?

  1. Sign up at codepeel.com.
  2. Click Connect GitHub and select your personal account or organization.
  3. Choose the repositories you want CodePeel to monitor.
  4. Open a pull request — CodePeel will review it automatically within 15–45 seconds.

Why did CodePeel skip my pull request?

If CodePeel did not review your PR, check the following common conditions:

  1. Draft PRs: Draft pull requests are ignored by default. Publish the PR to trigger a review.
  2. Title Filter: The PR title contains WIP, [draft], or DO NOT REVIEW (configured via auto_review.ignore_title_keywords).
  3. Target Branch: The PR does not target a configured base branch (default: main, master, develop).
  4. Bot Authors: PRs created by automated bots (e.g., Dependabot, Renovate) are skipped unless configured.
  5. Quota: Your monthly review allowance was reached (check Dashboard Billing).

How do I exclude generated files, lockfiles, or assets from review?

Add an ignore_paths block to .codepeel.yml in your repository root:

ignore_paths:
  - "node_modules/**"
  - "dist/**"
  - "build/**"
  - "*.lock"
  - "**/*.min.js"
  - "generated/**"

6. CodePeel vs. Copilot vs. Human Review

How is CodePeel different from GitHub Copilot or Cursor?

FeatureGitHub Copilot / CursorCodePeel
Primary RoleIn-editor code generation while typingAutomated senior reviewer at the CI/CD gate
Workflow TimingPre-commit (while writing)Pre-merge (pull requests & team review)
Security & SASTBasic autocomplete suggestionsDeep multi-layer SAST & secret scanning
Self-HealingRequires manual prompt fixingAutomatically generates ready-to-merge fix PRs
Team StandardsGeneric coding modelsEnforces custom regex rules and repo learnings

Does CodePeel replace human code reviewers?

No. CodePeel is designed to augment human reviewers, not replace them. CodePeel handles the tedious, time-consuming parts of code review — catching syntax traps, race conditions, missing error handlers, security leaks, and untested edge cases. This frees human reviewers to focus on business logic, product requirements, and system design.

7. Customization, Ecosystem & Integrations

Can I review code before opening a PR?

Yes! CodePeel supports pre-commit review across multiple developer surfaces:

  • VS Code Extension: Install the official CodePeel extension to review staged changes directly in your editor before pushing.
  • MCP Server (Model Context Protocol): Connect CodePeel directly to AI coding agents like Claude Code, Cursor, Cline, and Windsurf. Your AI coding agent can call review_code on its own work before presenting it to you.

What languages and frameworks are supported?

CodePeel supports all major programming languages out of the box, including:

  • Languages: TypeScript, JavaScript, Python, Go, Rust, Java, C#, C/C++, PHP, Ruby, Swift, Kotlin, Dart/Flutter, Shell, and SQL.
  • Test Frameworks: Jest, Vitest, PyTest, Go Test, Cargo Test, JUnit, and Flutter Test (for automated unit test generation).

Getting Help & Support

← All docsCodePeel