Skip to content

[Resource]: selvage — LLM-based code review MCP server with AST-powered smart context extraction #760

@anomie7

Description

@anomie7

Resource Submission

Name: selvage

URL: https://github.com/selvage-lab/selvage

Category: MCP Servers

Description:
selvage is an open-source LLM-based code review MCP server. It integrates directly with Claude Code (and other agents) to perform AI-powered code reviews on your git diffs.

Key Features:

  • get_review_context tool — agent-delegated review mode: returns the full diff + AST-extracted smart context as a structured prompt, letting Claude Code perform the review using its own LLM (no API key required)
  • review tool — independent review mode using your configured LLM (Claude, GPT, Gemini, 20+ models via OpenRouter)
  • AST-powered smart context extraction via tree-sitter — automatically pulls in related functions, classes, and imports beyond the raw diff
  • Supports staged, unstaged, branch-diff, and commit-range review modes

Why it's relevant to Claude Code users:
selvage is purpose-built as a Claude Code plugin. The get_review_context tool is designed for agent delegation — Claude Code calls it, receives the structured prompt, and performs the review itself. Zero extra API keys needed.

Install:

pip install selvage
selvage mcp install  # auto-installs into Claude Code's MCP config

Metadata

Metadata

Assignees

No one assigned

    Labels

    needs-templateIssue is a resource recommendation that did not use the issue template

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions