Display Name
selvage
Category
Tooling
Sub-Category
Slash-Commands: Code Analysis & Testing
Primary Link
https://github.com/selvage-lab/selvage
Author Name
DenimCoder
Author Link
https://github.com/anomie7
License
Apache-2.0
Other License
No response
Description
An MCP server that performs AI-powered code reviews on git diffs with AST-powered smart context extraction via tree-sitter.
Supports two modes: agent-delegated review
(get_review_context returns the structured prompt for Claude Code to review using its own LLM, no extra API key required) and
independent review using a configured LLM (Claude, GPT, Gemini, or 20+ models via OpenRouter).
Validate Claims
Install selvage and run selvage mcp install to register it in Claude Code.
Stage some changes in a git repo (git add -p), then ask Claude Code to review them using selvage.
Claude Code will call get_review_context, receive the diff with smart context, and return a structured code review — all without any additional API key.
Specific Task(s)
Ask Claude Code to perform a code review on your staged git changes using selvage.
Specific Prompt(s)
"Use selvage to review my staged changes and give me a detailed code review."
Additional Comments
selvage mcp install auto-configures Claude Code's MCP settings.
The get_review_context tool is specifically designed for agent delegation —
Claude Code calls it, receives the diff + AST-extracted context, and performs the review itself. Network requests: only made when using independent review mode (calls configured LLM provider APIs).
In agent-delegated mode, no external network requests are made by selvage.
Recommendation Checklist
Display Name
selvage
Category
Tooling
Sub-Category
Slash-Commands: Code Analysis & Testing
Primary Link
https://github.com/selvage-lab/selvage
Author Name
DenimCoder
Author Link
https://github.com/anomie7
License
Apache-2.0
Other License
No response
Description
An MCP server that performs AI-powered code reviews on git diffs with AST-powered smart context extraction via tree-sitter.
Supports two modes: agent-delegated review
(get_review_context returns the structured prompt for Claude Code to review using its own LLM, no extra API key required) and
independent review using a configured LLM (Claude, GPT, Gemini, or 20+ models via OpenRouter).
Validate Claims
Install selvage and run selvage mcp install to register it in Claude Code.
Stage some changes in a git repo (git add -p), then ask Claude Code to review them using selvage.
Claude Code will call get_review_context, receive the diff with smart context, and return a structured code review — all without any additional API key.
Specific Task(s)
Ask Claude Code to perform a code review on your staged git changes using selvage.
Specific Prompt(s)
"Use selvage to review my staged changes and give me a detailed code review."
Additional Comments
selvage mcp install auto-configures Claude Code's MCP settings.
The get_review_context tool is specifically designed for agent delegation —
Claude Code calls it, receives the diff + AST-extracted context, and performs the review itself. Network requests: only made when using independent review mode (calls configured LLM provider APIs).
In agent-delegated mode, no external network requests are made by selvage.
Recommendation Checklist