Mule AI v0.1.7: The Agent That Writes Its Own Pull Requests

What’s the most tedious part of software development? For me, it’s writing boilerplate code and creating pull requests. That’s why I’m excited to share what’s new in Mule AI v0.1.7 - the Implement Phase - where the agent can now write code AND create the PR for you.

The Problem: Context Switching is Expensive

We’ve all been there. You have a clear idea, you explain it to your AI assistant, it generates the code… and then you have to:

1. Copy the code to your local environment
2. Create a branch
3. Commit the changes
4. Push and create a PR
5. Fill in the description

It’s not rocket science, but it’s friction. And friction is the enemy of productivity.

Enter: The Implement Phase

Starting with v0.1.7, Mule can now handle the entire implementation workflow:

phases:
  - implement:
      module: wasm
      action: implement
      prompt: "Create a function that calculates Fibonacci numbers recursively with memoization"

The WASM module handles:

• Code Generation: Based on your specifications
• File Operations: Creates/updates files in the repository
• Git Operations: Creates branches, commits, and pushes
• PR Creation: Opens a pull request with your description

How It Works

The magic happens through a new WASM module that interfaces with Mule’s engine. Here’s what happens when you trigger the implement phase:

1. Analysis: The agent analyzes your prompt and determines what files need to be created or modified
2. Generation: Code is generated based on best practices and project conventions
3. Validation: Basic syntax and structure checks
4. Implementation: Files are written to a new branch
5. PR Creation: A pull request is opened with a descriptive summary

A Real Example

Recently, I used this feature to add a new bash tool to Mule. Here’s how it went:

User: "Add a bash execution tool that supports timeout and streaming output"

Mule (internally):
→ Analyzed existing tool patterns
→ Generated bash_tool.go with proper error handling
→ Added configuration for timeout settings
→ Created unit tests
→ Committed to feature/bash-tool branch
→ Opened PR #95

The result? A fully functional bash tool with streaming support, timeout handling, and comprehensive tests - all from a single prompt.

Why This Matters

This isn’t just about convenience. The Implement Phase represents a shift in how we think about AI assistants:

• Autonomy: Instead of just suggesting code, Mule can now deliver working implementations
• Traceability: Every change comes through proper code review workflows
• Consistency: Implementation follows project conventions automatically

Looking Forward

The v0.1.7 release is just the beginning. There’s already an open issue (#101) about updating the agent runtime to use pi - which opens up even more possibilities for enhanced reasoning and tool use.

The future of AI-assisted development isn’t about replacing developers - it’s about removing the mechanical barriers so we can focus on the creative parts. The Implement Phase is a significant step in that direction.

Want to try it out? Head over to github.com/mule-ai/mule and upgrade to v0.1.7. As always, I’m curious to hear your thoughts and see what you build!