diff --git a/README.md b/README.md
index 8266829..40c5aaa 100644
--- a/README.md
+++ b/README.md
@@ -1 +1,64 @@
-# code-review-agent
+# Code Review Agent
+
+AI-powered Python code review agent built with LangGraph and FastAPI.
+Analyzes Python code for typing issues, code quality, and security vulnerabilities.
+
+## What it does
+
+Accepts Python code via REST API and returns a structured review report with:
+- **Score** (0–10)
+- **Issues** — typing, quality, security, structure with severity levels
+- **Suggestions** — concrete improvements
+- **Summary** — overall assessment
+
+Non-Python code is rejected automatically.
+
+## Agent flow
+```
+check_if_python → analyze_structure → check_typing → check_quality → check_security → generate_report
+```
+
+## Project structure
+```
+├── app/
+│   ├── agent/
+│   │   ├── graph.py      # LangGraph agent definition
+│   │   └── tools.py
+│   ├── main.py           # FastAPI endpoints
+│   └── schemas.py        # Pydantic models
+├── requirements.txt
+└── README.md
+```
+
+## Setup
+```bash
+python -m venv venv
+source venv/bin/activate
+pip install -r requirements.txt
+```
+
+Create `.env` file in project root:
+```
+GOOGLE_API_KEY=your_key_here
+```
+
+## Run
+```bash
+uvicorn app.main:app --reload
+```
+
+API docs available at `http://127.0.0.1:8000/docs`
+
+## Usage
+```bash
+curl -X POST http://127.0.0.1:8000/review \
+  -H "Content-Type: application/json" \
+  -d '{"code": "def add(a, b):\n    return a + b"}'
+```
+
+## Limitations
+
+- Python only, max 5000 characters
+- 5 sequential LLM calls per request — slow on large inputs
+- Depends on Gemini API quota
+- LLM may hallucinate issues
diff --git a/app/__pycache__/__init__.cpython-314.pyc b/app/__pycache__/__init__.cpython-314.pyc
new file mode 100644
index 0000000..28ffd7d
Binary files /dev/null and b/app/__pycache__/__init__.cpython-314.pyc differ
diff --git a/app/__pycache__/main.cpython-314.pyc b/app/__pycache__/main.cpython-314.pyc
new file mode 100644
index 0000000..7728699
Binary files /dev/null and b/app/__pycache__/main.cpython-314.pyc differ
diff --git a/app/__pycache__/schemas.cpython-314.pyc b/app/__pycache__/schemas.cpython-314.pyc
new file mode 100644
index 0000000..ddc0ffa
Binary files /dev/null and b/app/__pycache__/schemas.cpython-314.pyc differ
diff --git a/app/agent/__pycache__/__init__.cpython-314.pyc b/app/agent/__pycache__/__init__.cpython-314.pyc
new file mode 100644
index 0000000..db2c9a2
Binary files /dev/null and b/app/agent/__pycache__/__init__.cpython-314.pyc differ
diff --git a/app/agent/__pycache__/graph.cpython-314.pyc b/app/agent/__pycache__/graph.cpython-314.pyc
new file mode 100644
index 0000000..272e897
Binary files /dev/null and b/app/agent/__pycache__/graph.cpython-314.pyc differ
diff --git a/app/agent/graph.py b/app/agent/graph.py
index e69de29..2968276 100644
--- a/app/agent/graph.py
+++ b/app/agent/graph.py
@@ -0,0 +1,204 @@
+from typing import TypedDict
+from langgraph.graph import StateGraph, END
+from langchain_google_genai import ChatGoogleGenerativeAI
+from dotenv import load_dotenv
+import json
+import re
+
+load_dotenv()
+
+llm = ChatGoogleGenerativeAI(model="gemini-2.5-flash")
+
+
+class ReviewState(TypedDict):
+    code: str
+    is_python: bool
+    structure: str
+    typing_issues: str
+    quality_issues: str
+    security_issues: str
+    report: dict
+
+
+def check_if_python(state: ReviewState) -> ReviewState:
+    prompt = """Is the following code written in Python? Reply with only "yes" or "no".
+    
+Code:
+{code}""".format(code=state["code"])
+    response = llm.invoke(prompt)
+    is_python = response.content.strip().lower().startswith("yes")
+    return {**state, "is_python": is_python}
+
+
+def reject(state: ReviewState) -> ReviewState:
+    return {
+        **state,
+        "report": {
+            "error": "Only Python code is supported.",
+            "issues": [],
+            "suggestions": [],
+            "score": None,
+        },
+    }
+
+
+def analyze_structure(state: ReviewState) -> ReviewState:
+    prompt = """Analyze the structure of this Python code.
+List: functions, classes, imports, and overall organization.
+Be concise.
+
+Code:
+{code}""".format(code=state["code"])
+    response = llm.invoke(prompt)
+    return {**state, "structure": response.content}
+
+
+def check_typing(state: ReviewState) -> ReviewState:
+    prompt = """Review the type annotations in this Python code.
+Check for:
+- Missing type hints on function arguments and return types
+- Incorrect use of Optional, Union, List, Dict etc.
+- Missing Pydantic model validation if applicable
+- Any type-unsafe patterns
+
+Code:
+{code}""".format(code=state["code"])
+    response = llm.invoke(prompt)
+    return {**state, "typing_issues": response.content}
+
+
+def check_quality(state: ReviewState) -> ReviewState:
+    prompt = """Review the code quality of this Python code.
+Check for:
+- Poor naming conventions
+- Missing or incomplete docstrings
+- Functions that are too complex or too long
+- Code duplication
+- Violations of PEP8
+
+Code:
+{code}""".format(code=state["code"])
+    response = llm.invoke(prompt)
+    return {**state, "quality_issues": response.content}
+
+
+def check_security(state: ReviewState) -> ReviewState:
+    prompt = """Review this Python code for security issues.
+Check for:
+- Use of eval() or exec()
+- SQL injection vulnerabilities
+- Files opened without context manager
+- Hardcoded secrets or passwords
+- Unsafe deserialization
+
+Code:
+{code}""".format(code=state["code"])
+    response = llm.invoke(prompt)
+    return {**state, "security_issues": response.content}
+
+
+def generate_report(state: ReviewState) -> ReviewState:
+    prompt = """You are a senior Python code reviewer.
+Based on the analysis below, generate a structured code review report.
+
+Structure analysis:
+{structure}
+
+Typing issues:
+{typing_issues}
+
+Code quality issues:
+{quality_issues}
+
+Security issues:
+{security_issues}
+
+Respond ONLY with a valid JSON object in this exact format, no markdown, no extra text:
+{{
+  "score": <integer from 0 to 10>,
+  "issues": [
+    {{"type": "typing|quality|security|structure", "severity": "low|medium|high", "description": "..."}}
+  ],
+  "suggestions": [
+    "suggestion 1",
+    "suggestion 2"
+  ],
+  "summary": "overall summary in 2-3 sentences"
+}}""".format(
+        structure=state["structure"],
+        typing_issues=state["typing_issues"],
+        quality_issues=state["quality_issues"],
+        security_issues=state["security_issues"],
+    )
+
+    response = llm.invoke(prompt)
+
+    raw = response.content.strip()
+    raw = re.sub(r"^```json\s*", "", raw)
+    raw = re.sub(r"```$", "", raw).strip()
+
+    try:
+        report = json.loads(raw)
+    except json.JSONDecodeError:
+        report = {"error": "Failed to parse report", "raw": raw}
+
+    return {**state, "report": report}
+
+
+def route_after_check(state: ReviewState) -> str:
+    if state["is_python"]:
+        return "analyze_structure"
+    return "reject"
+
+
+def build_graph():
+    graph = StateGraph(ReviewState)
+
+    graph.add_node("check_if_python", check_if_python)
+    graph.add_node("reject", reject)
+    graph.add_node("analyze_structure", analyze_structure)
+    graph.add_node("check_typing", check_typing)
+    graph.add_node("check_quality", check_quality)
+    graph.add_node("check_security", check_security)
+    graph.add_node("generate_report", generate_report)
+
+    graph.set_entry_point("check_if_python")
+
+    graph.add_conditional_edges(
+        "check_if_python",
+        route_after_check,
+        {"analyze_structure": "analyze_structure", "reject": "reject"},
+    )
+
+    graph.add_edge("reject", END)
+    graph.add_edge("analyze_structure", "check_typing")
+    graph.add_edge("check_typing", "check_quality")
+    graph.add_edge("check_quality", "check_security")
+    graph.add_edge("check_security", "generate_report")
+    graph.add_edge("generate_report", END)
+
+    return graph.compile()
+
+
+if __name__ == "__main__":
+    agent = build_graph()
+    test_code = """
+def add(a, b):
+    return a + b
+
+class Calculator:
+    def multiply(self, x, y):
+        return x * y
+"""
+    result = agent.invoke(
+        {
+            "code": test_code,
+            "is_python": False,
+            "structure": "",
+            "typing_issues": "",
+            "quality_issues": "",
+            "security_issues": "",
+            "report": {},
+        }
+    )
+    print(json.dumps(result["report"], indent=2))
diff --git a/app/main.py b/app/main.py
index e69de29..a2ca119 100644
--- a/app/main.py
+++ b/app/main.py
@@ -0,0 +1,22 @@
+from fastapi import FastAPI
+from app.schemas import ReviewRequest, ReviewResponse
+from app.agent.graph import build_graph
+
+app = FastAPI(title="Python Code Review Agent")
+graph = build_graph()
+
+
+@app.post("/review", response_model=ReviewResponse)
+def review_code(request: ReviewRequest):
+    result = graph.invoke(
+        {
+            "code": request.code,
+            "is_python": False,
+            "structure": "",
+            "typing_issues": "",
+            "quality_issues": "",
+            "security_issues": "",
+            "report": {},
+        }
+    )
+    return ReviewResponse(**result["report"])
diff --git a/app/schemas.py b/app/schemas.py
index e69de29..11d17fb 100644
--- a/app/schemas.py
+++ b/app/schemas.py
@@ -0,0 +1,20 @@
+from pydantic import BaseModel
+from typing import Optional
+
+
+class ReviewRequest(BaseModel):
+    code: str
+
+
+class ReviewIssue(BaseModel):
+    type: str
+    severity: str
+    description: str
+
+
+class ReviewResponse(BaseModel):
+    score: Optional[int] = None
+    issues: list[ReviewIssue] = []
+    suggestions: list[str] = []
+    summary: Optional[str] = None
+    error: Optional[str] = None
diff --git a/requirements.txt b/requirements.txt
index e69de29..e9ad4e8 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -0,0 +1,6 @@
+fastapi
+langgraph
+langchain-google-genai
+langchain-core
+uvicorn
+python-dotenv