Code Reviewer LangGraph

blueprint · 4 agents · 1 MCP server

1300★ stars

210forks

4.8rating

What it does

A diff-aware reviewer with a built-in skeptic. A reader pulls the pull request, a finder proposes issues, a skeptic tries to refute each one, and a reporter writes up only the findings that survived. Fewer false positives than a single pass.

The cast · 4 agents

Finderfinderreasoning

Read the diff and propose specific findings: bugs, risks, and smells.

no tools · reasoning only

Readerreaderfast

Fetch the pull request and its diff.

gh.get_pr

Skepticskepticreasoning

Try to refute each finding. Drop the ones that don't hold up.

no tools · reasoning only

Reporterreporterreasoning

Write up the surviving findings as a review comment.

gh.comment

Flow

→ Reader → diff

→ Finder → findings

→ Skeptic → confirmed

→ Reporter → review

Interface

Inputs

prstring

The pull request to review.

Outputs

reviewstring

The posted review.

Depends on · 1 MCP server

io.github.github/github-mcpnot in the registry yetgh

⤓ Export runnable code

# Generated by @socketcat/compiler for target: langgraph
# blueprint: com.socketcat/code-reviewer v1.0.0   schema: socketcat.dev/blueprint/v0
# This code is yours. Edit it freely. The socketcat_runtime helper is optional and can be vendored.

from typing import TypedDict
from langgraph.graph import StateGraph, START, END
import socketcat_runtime as _rt

class State(TypedDict, total=False):
    pr: object
    review: object
    diff: object
    findings: object
    confirmed: object

def node_AgentInvoke_0(state):
    return _rt.run_agent("reader", state, {"pr":"pr"}, "diff")

def node_AgentInvoke_1(state):
    return _rt.run_agent("finder", state, {"diff":"diff"}, "findings")

def node_AgentInvoke_2(state):
    return _rt.run_agent("skeptic", state, {"findings":"findings"}, "confirmed")

def node_AgentInvoke_3(state):
    return _rt.run_agent("reporter", state, {"confirmed":"confirmed"}, "review")

def build():
    b = StateGraph(State)
    b.add_node("AgentInvoke_0", node_AgentInvoke_0)
    b.add_node("AgentInvoke_1", node_AgentInvoke_1)
    b.add_node("AgentInvoke_2", node_AgentInvoke_2)
    b.add_node("AgentInvoke_3", node_AgentInvoke_3)
    b.add_edge(START, "AgentInvoke_0")
    b.add_edge("AgentInvoke_0", "AgentInvoke_1")
    b.add_edge("AgentInvoke_1", "AgentInvoke_2")
    b.add_edge("AgentInvoke_2", "AgentInvoke_3")
    b.add_edge("AgentInvoke_3", END)
    return b.compile()

INPUTS = ["pr"]
OUTPUTS = ["review"]

if __name__ == "__main__":
    _rt.main(build, INPUTS, OUTPUTS)

▸blueprint.json (the portable format)

{
  "id": "com.socketcat/code-reviewer",
  "flow": {
    "type": "sequence",
    "blocks": [
      {
        "in": {
          "pr": "pr"
        },
        "out": "diff",
        "use": "reader",
        "type": "agent"
      },
      {
        "in": {
          "diff": "diff"
        },
        "out": "findings",
        "use": "finder",
        "type": "agent"
      },
      {
        "in": {
          "findings": "findings"
        },
        "out": "confirmed",
        "use": "skeptic",
        "type": "agent"
      },
      {
        "in": {
          "confirmed": "confirmed"
        },
        "out": "review",
        "use": "reporter",
        "type": "agent"
      }
    ]
  },
  "tags": [
    "code-review",
    "verification"
  ],
  "title": "Code Reviewer",
  "agents": {
    "finder": {
      "model": {
        "hint": "reasoning"
      },
      "title": "Finder",
      "instructions": "Read the diff and propose specific findings: bugs, risks, and smells."
    },
    "reader": {
      "model": {
        "hint": "fast"
      },
      "title": "Reader",
      "tools": [
        "gh.get_pr"
      ],
      "instructions": "Fetch the pull request and its diff."
    },
    "skeptic": {
      "model": {
        "hint": "reasoning"
      },
      "title": "Skeptic",
      "instructions": "Try to refute each finding. Drop the ones that don't hold up."
    },
    "reporter": {
      "model": {
        "hint": "reasoning"
      },
      "title": "Reporter",
      "tools": [
        "gh.comment"
      ],
      "instructions": "Write up the surviving findings as a review comment."
    }
  },
  "estate": {
    "diff": {
      "type": "object",
      "description": "The pull request diff and metadata the reader fetched."
    },
    "review": {
      "type": "string",
      "description": "The review comment the reporter wrote up."
    },
    "findings": {
      "type": "object",
      "description": "Issues the finder proposed, before verification."
    },
    "confirmed": {
      "type": "object",
      "description": "The findings that survived the skeptic's refutation pass."
    }
  },
  "$schema": "socketcat.dev/blueprint/v0",
  "authors": [
    {
      "url": "https://socketcat.com",
      "name": "SocketCat"
    }
  ],
  "license": "MIT",
  "servers": [
    {
      "ref": "io.github.github/github-mcp",
      "alias": "gh"
    }
  ],
  "summary": "Reads a diff, proposes findings, verifies each with a skeptic, and reports only survivors.",
  "targets": [
    "langgraph",
    "crewai",
    "*"
  ],
  "version": "1.0.0",
  "interface": {
    "inputs": {
      "pr": {
        "type": "string",
        "description": "The pull request to review."
      }
    },
    "outputs": {
      "review": {
        "type": "string",
        "description": "The posted review."
      }
    }
  },
  "description": "A diff-aware reviewer with a built-in skeptic. A reader pulls the pull request, a finder proposes issues, a skeptic tries to refute each one, and a reporter writes up only the findings that survived. Fewer false positives than a single pass."
}