aicraft-code-review

MCP Code Review Server

mcp-name: io.github.GoodJobwilliam/aicraft-code-review

中文文档：README.zh.md

Listed on: Smithery · mcpservers.org · cursor.directory

Code review as an MCP server. Connect it to Claude Code, Cursor, or any MCP-compatible AI assistant.

Features

• review_code — Review any source code snippet for bugs, security, performance, and style
• review_diff — Review a git diff for potential issues before merging
• review_file — Review a local file by path

Powered by the same methodology as our Code Review Agent: OWASP Top 10 scanning, N+1 query detection, race condition analysis, and structured output with severity ratings.

Quick Start

Via uvx (no install)

# Add to your Claude Code MCP config:
claude mcp add code-review -- uvx --with "mcp<2" aicraft-code-review

Or add to your ~/.cursor/mcp.json or claude_desktop_config.json:

{
  "mcpServers": {
    "code-review": {
      "command": "uvx",
      "args": ["--with", "mcp<2", "aicraft-code-review"]
    }
  }
}

Via pip

pip install "aicraft-code-review" "mcp<2"
python -m mcp_code_review

⚠️ PyPI 0.1.0 does not cap mcp, and mcp 2.0.0 removes Server.list_tools, which crashes the CLI on fresh installs. Install 0.1.1+ (which pins mcp<2 automatically) or pin mcp<2 yourself.

CLI mode (no MCP client needed)

# Review a local file (discovers .mcp-code-review.yaml from the file's directory)
mcp-code-review review-file path/to/file.py

# Review a git diff
git diff | mcp-code-review review-diff
mcp-code-review review-diff --git

# Review a snippet
mcp-code-review review-code "import os; os.system('ls')"

Exit codes are CI-friendly: 0 clean, 1 high/medium issues, 2 critical issues.

Usage Examples

Once connected, ask your AI assistant:

"Review this Python code for security issues: [paste code]" "Review this diff before I commit: [paste diff]" "Review this file: /path/to/file.py"

The AI will call the MCP server and return structured results.

Sample Output

## Review Results

### 🔴 Critical (1)
| Line | Issue | Category | Fix |
|------|-------|----------|-----|
| 42 | SQL injection via f-string | Security | Use parameterized queries |

### 🟠 High (2)
| Line | Issue | Category | Fix |
|------|-------|----------|-----|
| 15 | Unvalidated user input | Security | Add input validation |
| 78 | N+1 query in loop | Performance | Add select_related |

### Summary
- **Critical**: 1 — must fix
- **High**: 2 — should fix
- **Medium**: 0
- **Info**: 0

Custom Rules & Team Profiles

Ship your team's code standards as a config file — no code changes needed.

• .mcp-code-review.yaml / .yml / .json — auto-discovered from the reviewed file's directory upward; for snippets and diffs it is looked up from the server's working directory
• MCP_CODE_REVIEW_CONFIG env var — point every teammate at a shared config committed to your repo (team-shared rule profiles)
• Custom regex rules with severity, message, and suggested fix
• disabled_checks — silence noisy checks
• severity_overrides — bump or lower any check (e.g. make hardcoded secrets blocking)
• min_severity — only report findings at or above a threshold (per-repo noise control)

Example .mcp-code-review.yaml

disabled_checks:
  - todo_comment

severity_overrides:
  hardcoded_secret: critical

min_severity: medium

custom_rules:
  - name: no-console-log
    pattern: 'console\.log\('
    severity: high
    category: quality
    issue: Console logging left in production code
    fix: Use a structured logger instead

Team setup

Commit the file to a shared repo, then wire every teammate's MCP client to it:

{
  "mcpServers": {
    "code-review": {
      "command": "uvx",
      "args": ["--with", "mcp<2", "aicraft-code-review"],
      "env": {
        "MCP_CODE_REVIEW_CONFIG": "/path/to/team-repo/.mcp-code-review.yaml"
      }
    }
  }
}

Available check ids: dynamic_exec, sql_injection, deserialization, command_injection, input_py2, xss_innerhtml, hardcoded_secret, nplus1, unbounded_list, bare_except, empty_except, todo_comment, missing_return_type, long_lines, snake_case, pascal_case.

YAML configs need pip install "aicraft-code-review[yaml]"; JSON configs work with no extra dependencies.

Development

git clone https://github.com/GoodJobwilliam/aicraft
cd aicraft
pip install -e ".[dev]"
python -m mcp_code_review  # Start server

Requirements

• Python 3.11+
• An MCP-compatible client (Claude Code, Cursor, etc.)

License

MIT