Claude-Codex Duo
Claude codes. Codex reviews. Better code together.
A simple library for duo peer programming between Claude Code and Codex. Two AI coding assistants working together to write better code.
Why Use Both?
| Claude | Codex |
|---|---|
| Superior tool calling & code generation | Excellent at catching bugs & edge cases |
| Great at understanding context | Strong code review capabilities |
| Creative problem solving | Thorough analysis |
Together, they create a peer programming experience that's greater than either alone.
Quick Start
Prerequisites
- Python 3.10+
- Claude Code CLI installed
- Codex CLI installed
Installation
pip install claude-codex-duo
Or install from source:
git clone https://github.com/Sigma5C-Corp/claude-codex-duo
cd claude-codex-duo
pip install -e .
Basic Usage
from claude_codex_duo import create_session, Verdict
# Create a peer programming session
session = create_session("auth_feature", "Implement user authentication")
# Claude submits code for review
session.submit_code("""
def authenticate(username: str, password: str) -> bool:
user = get_user(username)
if user and verify_password(password, user.password_hash):
return True
return False
""")
# Wait for Codex's review
exchange = session.wait_for_review(timeout=120)
if exchange and exchange.verdict == Verdict.APPROVED:
print("Consensus reached! Code approved.")
else:
print(f"Feedback: {exchange.codex_review}")
How It Works
┌─────────────────────────────────────────────────────────────────┐
│ │
│ ┌──────────┐ ┌──────────┐ │
│ │ Claude │ ───── submits code ────► │ Codex │ │
│ │ (coder) │ │(reviewer)│ │
│ └──────────┘ ◄──── sends review ───── └──────────┘ │
│ │ │ │
│ │ ┌─────────────┐ │ │
│ └─────────►│ Session │◄─────────────┘ │
│ │ (files) │ │
│ └─────────────┘ │
│ │
│ Round 1: Claude writes → Codex reviews → Feedback │
│ Round 2: Claude fixes → Codex reviews → Approved! │
│ │
└─────────────────────────────────────────────────────────────────┘
- Claude writes code - Implements the task
- Codex reviews - Checks for bugs, edge cases, best practices
- Iterate - Claude addresses feedback, Codex re-reviews
- Consensus - When Codex approves, the code is ready
All communication happens via file-based messaging (no database needed).
API Reference
create_session(session_id, task, workspace=None, max_rounds=10)
Create a new peer programming session.
session = create_session(
"my_session", # Unique session ID
"Build a REST API", # Task description
workspace="/path/to/project",
max_rounds=5
)
DuoSession.submit_code(content, files_modified=None)
Submit code from Claude for Codex review.
session.submit_code(
"def hello(): return 'world'",
files_modified=["src/utils.py"]
)
DuoSession.wait_for_review(timeout=300)
Wait for Codex's review response.
exchange = session.wait_for_review(timeout=120)
if exchange:
print(f"Verdict: {exchange.verdict}")
print(f"Feedback: {exchange.codex_review}")
Session Status
session.status # SessionStatus.ACTIVE, REVIEWING, CONSENSUS, etc.
session.has_consensus # True if approved
session.is_complete # True if done (approved or max rounds)
session.current_round # Current review round number
Configuration
Set the data directory via environment variable:
export CLAUDE_CODEX_DUO_DATA=~/.my-duo-data
Default: ~/.claude-codex-duo/
Session files are stored in:
~/.claude-codex-duo/
└── sessions/
└── my_session/
├── session.json # Session state
├── claude_inbox.json # Messages for Claude
├── codex_inbox.json # Messages for Codex
└── conversation.jsonl # Full history
Advanced Usage
Manual Provider Control
from claude_codex_duo import ClaudeProvider, CodexProvider
# Direct Claude interaction
claude = ClaudeProvider(workspace="/my/project")
response = claude.execute("Write a function to validate emails")
# Direct Codex interaction
codex = CodexProvider(model="o3", reasoning_effort="high")
review = codex.execute(codex.review_prompt(code, task))
verdict = codex.extract_verdict(review)
Load Existing Session
from claude_codex_duo import load_session
session = load_session("my_session")
print(f"Round {session.current_round} of {session.max_rounds}")
Requirements
- Python 3.10+
- Unix-like OS (uses
fcntlfor file locking) - Claude Code CLI
- Codex CLI
License
MIT License - see LICENSE
Contributing
Contributions welcome! Please read our contributing guidelines.
About
Built by Sigma5C - Leaders in multi-agent AI systems.
We believe AI tools work better together. Claude's superior reasoning and tool-calling combined with Codex's bug-catching creates a peer programming experience that's greater than either alone.
"Two AIs are better than one."
Metadata
Release files for claude-codex-duo 0.1.1
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| claude_codex_duo-0.1.1.tar.gz | 12.9 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| claude_codex_duo-0.1.1-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 26.9 kB
Release files / claude_codex_duo-0.1.1.tar.gz
| Download URL | claude_codex_duo-0.1.1.tar.gz |
|---|---|
| Size | 12.9 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
6f3256efedc947e3fd60d3492ebfca48ae99fabe664526ed4bd9e6fba7491f3b
|
|
BLAKE2b-256 checksum How to use checksums |
91f82df27d0f75e2dd06d81e353e03df919610feb62711f84e5acade69b315b1
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jan 19, 2026.
Transparency logRelease files / claude_codex_duo-0.1.1-py3-none-any.whl
| Download URL | claude_codex_duo-0.1.1-py3-none-any.whl |
|---|---|
| Size | 14.0 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
2a2d26cb19a2147c70e584df9a38b22df65d13c4c608c234cf57f94fb9d44b30
|
|
BLAKE2b-256 checksum How to use checksums |
d396ddbd9bf7f848089bd64d54f9fbcd70eab38ee3ce8dafcaa84116b1afa756
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.13.7
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Jan 19, 2026.
Transparency log