Code Sage

The intelligent development companion.

Key Features

Security Analysis
Detects vulnerabilities, injection risks, and insecure patterns.
Maintainability
Scores complexity, coupling, and tech debt to guide refactors.
Bug Detection
Surfaces logical errors and edge-case failures before they ship.
Detects Hallucination
Flags AI-generated code that looks plausible but hides subtle errors.

Ready to dive in?

Full API reference, CLI flags, provider configuration, and real-world usage examples — all in one place.

MIT License — Free & Open Source Star on GitHub ↗

Quick Start

git clone https://github.com/nalintiwary18/codesage
cd codesage
pip install -e .
codesage run /path/to/project \
  --no-wizard \
  --provider openai \
  --model gpt-4o-mini \
  --output report.md

Code Sage

The intelligent development companion.

Key Features

Security Analysis

Detects vulnerabilities, injection risks, and insecure patterns.

Maintainability

Scores complexity, coupling, and tech debt to guide refactors.

Bug Detection

Surfaces logical errors and edge-case failures before they ship.

Detects Hallucination

Flags AI-generated code that looks plausible but hides subtle errors.

Ready to dive in?

Full API reference, CLI flags, provider configuration, and real-world usage examples — all in one place.

MIT License View on GitHub

Quick Start

git clone ...codesage
cd codesage
pip install -e .
codesage run .

CLI Reference · Python 3.11+

Documentation

openaigeminianthropicgroqollama

codesage run [TARGET] · all flags

Flag	Value	Description
--provider, -p	openai\|gemini\|anthropic\|groq\|ollama	LLM provider to use
--model, -m	<model-id>	Specific model name
--output, -o	<file.md>	Output report path
--max-chunks	<int>	Max source chunks to analyse
--no-wizard	—	Skip interactive setup
--no-cache	—	Bypass result cache
--verbose, -v	—	Enable verbose logging

codesage init [TARGET]

Generate .codesage.yml interactively

codesage doctor

Check Python, deps & API keys

Common usage patterns(Scroll for more)

# Install
pip install -e .

# Analyse current dir (wizard)
codesage run

# Analyse a path, no wizard
codesage run ./my-project \
  --no-wizard \
  --provider openai \
  --model gpt-4o-mini \
  --output report.md

# Larger analysis budget
codesage run . --max-chunks 100

# Init config for a project
codesage init ./my-project

# Health check
codesage doctor

Environment variables

OPENAI_API_KEYOpenAI key

GEMINI_API_KEYGoogle Gemini key

ANTHROPIC_API_KEYAnthropic key

GROQ_API_KEYGroq key

Config priority: CLI flags → env vars → .codesage.yml → defaults

Output

Reports written to ./reports/report.md by default. If the file exists, a timestamped copy is created instead of overwriting.

CLI Reference · Python 3.11+

Documentation

openaigeminianthropicgroqollama

codesage run [TARGET] · all flags

--provider, -popenai|gemini|anthropic|groq|ollama

LLM provider to use

--model, -m<model-id>

Specific model name

--output, -o<file.md>

Output report path

--max-chunks<int>

Max source chunks to analyse

--no-wizard—

Skip interactive setup

--no-cache—

Bypass result cache

--verbose, -v—

Enable verbose logging