Package Exports

@codepeelai/mcp-server
@codepeelai/mcp-server/dist/index.js

This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (@codepeelai/mcp-server) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

Readme

npm version Node.js version

CodePeel MCP Server

Finds the bugs. Opens the fix.

AI code review tools and prompts for Claude Code, Kiro, Cline, Roo, Cursor, and other MCP-compatible agents.
Same review engine that runs on 527+ GitHub PRs — 2,143 issues found in 29s average.

Why

When you're vibe coding with an AI agent, CodePeel acts as a second opinion — reviewing the generated code for bugs, security issues, and bad patterns before you commit.

The AI agent can even review its own output:

You: Build a login form with validation
Agent: [writes code, then calls review_code on its own output]
Agent: I found a potential XSS issue in my implementation. Let me fix that...
Agent: [calls fix_code, applies the fix]
Agent: Here's the login form — I caught and fixed a security issue during generation.

Tools

Tool	Description
`review_code`	Review a diff for bugs, security issues, and best practices
`fix_code`	Generate a fix for a specific code issue
`ask_codepeel`	Ask questions about code patterns, architecture, etc.
`check_credits`	Check your account balance and usage (free — does not consume a review)

`review_code`

Parameter	Type	Required	Description
`diff`	string	yes	Unified diff. Must be ≥ 10 characters and ≤ 500 KB.
`repo`	string	no	Repo name for context. Defaults to `"local"`. Accepts `"name"` or `"owner/name"`.

Consumes 1 review. 90s timeout.

`fix_code`

Parameter	Type	Required	Description
`file`	string	yes	File path (e.g. `"src/auth.ts"`)
`issue`	string	yes	Description of the problem
`problemCode`	string	no	The problematic snippet
`line`	number	no	Line number
`severity`	enum	no	`critical` / `high` / `medium` / `low`. Default `medium`.

Consumes 1 review. 60s timeout.

`ask_codepeel`

Parameter	Type	Required	Description
`question`	string	yes	Your question about the code
`diff`	string	no	Optional code context (a diff, file contents, or snippet)

Consumes 1 review. 60s timeout.

Prompts

Reusable workflows the agent can load to guide a review session. Prompts are free — only the tool calls they trigger consume reviews.

Prompt	Description
`review-staged-changes`	Pre-commit review: runs `git diff --cached` → `review_code` → reports findings
`security-audit`	Security-focused review with priority on injection, secrets, auth, and unsafe data handling
`explain-and-fix`	Full review cycle: review → explain each finding → generate fixes → apply. Accepts optional `language` argument.

Setup

1. Get your API token

Sign in at codepeel.com
Go to Settings → API Tokens
Click Create Token and copy the token (starts with cpk_)

API tokens don't expire — they work until you revoke them.

2. Configure your editor

Pick your editor and add the config block:

Claude Desktop

File: claude_desktop_config.json

{
  "mcpServers": {
    "codepeel": {
      "command": "npx",
      "args": ["-y", "@codepeelai/mcp-server"],
      "env": {
        "CODEPEEL_TOKEN": "cpk_your_token_here"
      }
    }
  }
}

Claude Code

File: .claude/settings.json

{
  "mcpServers": {
    "codepeel": {
      "command": "npx",
      "args": ["-y", "@codepeelai/mcp-server"],
      "env": {
        "CODEPEEL_TOKEN": "cpk_your_token_here"
      }
    }
  }
}

Kiro

File: .kiro/settings/mcp.json

{
  "mcpServers": {
    "codepeel": {
      "command": "npx",
      "args": ["-y", "@codepeelai/mcp-server"],
      "env": {
        "CODEPEEL_TOKEN": "cpk_your_token_here"
      }
    }
  }
}

Cline

In Cline settings, add an MCP server:

Command: npx
Args: -y @codepeelai/mcp-server
Env: CODEPEEL_TOKEN=cpk_your_token_here

Roo Code

In Roo Code settings (⚙️), add MCP server config:

{
  "mcpServers": {
    "codepeel": {
      "command": "npx",
      "args": ["-y", "@codepeelai/mcp-server"],
      "env": {
        "CODEPEEL_TOKEN": "cpk_your_token_here"
      }
    }
  }
}

Cursor

In Cursor settings → MCP, add:

{
  "codepeel": {
    "command": "npx",
    "args": ["-y", "@codepeelai/mcp-server"],
    "env": {
      "CODEPEEL_TOKEN": "cpk_your_token_here"
    }
  }
}

3. Use it

The AI agent will automatically have access to CodePeel's review tools and prompts. Ask it to:

"Review my recent changes for bugs"
"Check this diff for security issues"
"Fix the null pointer issue in auth.ts"
"How many reviews do I have left?"
Use the review-staged-changes prompt before committing

CLI

The package includes a CLI for interactive setup and standalone use:

npx -y @codepeelai/mcp-server init          # Interactive editor setup
npx -y @codepeelai/mcp-server init-config   # Generate .codepeel.yml from dashboard settings
npx -y @codepeelai/mcp-server credits       # Check your balance and plan
git diff | npx -y @codepeelai/mcp-server review  # Review a diff from stdin

When run with no arguments, starts the MCP server on stdio (used by AI editors).

Environment Variables

Variable	Required	Description
`CODEPEEL_TOKEN`	Yes	Your API token from Settings (starts with `cpk_`)
`CODEPEEL_API_URL`	No	API base URL (default: `https://www.codepeel.com/api`)
`CODEPEEL_REVIEW_URL`	No	Review endpoint URL (default: `CODEPEEL_API_URL` + `/ide/review`)

Staging

To use the MCP server against staging, set:

export CODEPEEL_API_URL=https://codepeel-staging.vercel.app/api
export CODEPEEL_TOKEN=cpk_your_staging_token

Or create a .env file in your project:

CODEPEEL_API_URL=https://codepeel-staging.vercel.app/api
CODEPEEL_REVIEW_URL=https://idereview-5ytww7s7ta-uc.a.run.app
CODEPEEL_TOKEN=cpk_your_staging_token

Troubleshooting

Error	Fix
`CODEPEEL_TOKEN not set`	Set the `CODEPEEL_TOKEN` environment variable in your MCP config (see Setup above).
"Token expired"	You're using a Firebase ID token instead of an API token. Create a `cpk_` token in Settings.
"Invalid token"	Check that your token starts with `cpk_` and hasn't been revoked.
"No credits remaining"	Upgrade your plan at codepeel.com/#pricing or wait for credits to reset.
"Rate limit exceeded"	Wait the indicated time and retry. Free: 3/hour, Pro: 6/hour, Max: 12/hour.
"Payment failed"	Update your payment method at codepeel.com/app/billing.
"Server error (5xx)"	CodePeel is temporarily unavailable. Try again in a moment.
Server starts but tools don't appear	Restart your AI editor after adding the MCP config. Verify the `command` is `"npx"` and `args` are `["-y", "@codepeelai/mcp-server"]`.

Full reference with response examples, edge cases, and workflow guides: codepeel.com/docs/mcp

License

MIT