Building Kodiak: An Open-Source Autonomous AI Software Engineer (Looking for Contributors)

Over the last few weeks, I’ve been building Kodiak, an open-source AI engineering platform with one long-term goal:

An autonomous AI software engineer capable of understanding a repository, planning engineering work, writing code, validating changes, and eventually creating production-ready pull requests.

This isn’t another chatbot.

It isn’t another wrapper around an LLM.

Kodiak is being built as an engineering system where multiple components work together to solve real software development problems.

The Vision

Imagine opening a repository and simply asking:

“Fix Issue #27.”

Instead of only generating code, the system would:

• Clone the repository
• Analyze the project structure
• Understand dependencies
• Build long-term memory
• Search documentation automatically
• Break the task into smaller engineering steps
• Choose the correct tools
• Generate code
• Execute tests
• Fix failures
• Learn from previous executions
• Request human approval when necessary
• Open a detailed pull request

That’s the direction Kodiak is heading.

• Current Tech Stack
• Python
• FastAPI
• PostgreSQL
• Redis
• ChromaDB
• Celery
• Docker
• SQLAlchemy
• Alembic
• Typer CLI
• Current Progress

So far Kodiak includes:

• Authentication
• Project management
• Repository management
• Task system
• Background workers
• REST API
• Memory foundation
• Docker deployment
• Database models
• CLI foundation

The project is still in active development, but the architecture is becoming much more modular.

What I’m Building Next

The next milestones include:

Repository indexing
Semantic code search
Multi-agent planning
Long-term memory
GitHub Issue execution
Autonomous code generation
Pull request generation
Plugin ecosystem
Local execution support
Looking for Contributors

I’m looking for developers interested in:

• AI Agents
• FastAPI
• Backend Engineering
• PostgreSQL
• SQLAlchemy
• Celery
• Docker
• CLI Development
• Testing
• Documentation Whether you’re an experienced engineer or someone wanting to contribute to open source, every contribution helps move the project forward.

I’d Love Feedback

*I’m especially interested in hearing:
*
What features would make you actually use a system like this?
What parts of the architecture would you redesign?
What problems do current AI coding tools still fail to solve?
If you were contributing, where would you start?

Constructive criticism is just as valuable as code contributions.

GitHub

Repository:
https://github.com/ShamGaneshan2008/Kodiak

If the vision sounds interesting, feel free to open an issue, submit a pull request, or simply share your thoughts.

Let’s build something the open-source AI community can genuinely use.