emmeongoingammuaroi/reviewform
AI-Powered Code Review Agent built with LangGraph, FastAPI, MCP, and RAG (Qdrant). Multi-step workflow with human-in-the-loop, eval framework, and structured logging.
Platform-specific configuration:
{
"mcpServers": {
"reviewform": {
"command": "npx",
"args": [
"-y",
"reviewform"
]
}
}
}Add the config above to .claude/settings.json under the mcpServers key.
A production-grade code review agent built with LangGraph, FastAPI, RAG (Qdrant), and MCP. It analyzes code snippets or GitHub PR diffs using a multi-step AI workflow, retrieves relevant coding standards via vector search, provides structured feedback, and supports human-in-the-loop refinement.
┌─────────────────────────────────────────────────────────────────┐
│ Clients │
│ (curl / Postman / Frontend / Claude Desktop / MCP clients) │
└──────────────┬──────────────────────────────┬────────────────────┘
│ HTTP/REST │ MCP (SSE or stdio)
▼ ▼
┌──────────────────────────┐ ┌───────────────────────────────┐
│ FastAPI Backend │ │ MCP Server (port 8001) │
│ (JWT auth, async) │ │ 2 tools: │
│ │ │ - fetch_github_pr_diff │
│ POST /reviews/ │ │ - search_coding_standards │
│ POST /reviews/{id}/ │ │ │
│ feedback │ │ Transports: │
│ GET /eval/logs │ │ - SSE (for LangGraph agent) │
│ POST /eval/score │ │ - stdio (for Claude Desktop) │
└──────────┬─────────────────┘ └──────────┬────────────────────┘
│ │
▼ │
┌──────────────────────────────────────────┐ │
│ LangGraph Review Workflow │ │
│ │ │
│ Agent nodes call tools via MCP client: │ │
│ │ │
│ ┌────────────┐ ┌──────────────────┐ │ │
│ │ fetch_diff │───▶│ retrieve_ │ │ │
│ │ (MCP call) │ │ standards │ │ │
│ └────────────┘ │ (MCP call → RAG) │ │ │
│ Loading reviews...