Skip to content

ManagementMO/CodeScribe

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

82 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

CodeScribe Logo

CodeScribe

From Linear ticket to pull request, automated.

An AI-powered agent ecosystem that automates the entire development pipeline, from generating initial code drafts to creating fully-packaged pull requests.

The ProblemThe SolutionHow It WorksQuick StartTech Stack

Hack the 6ix 2025 Badge Warp Badge Node.js Badge Google Gemini Badge Vellum Badge


The Problem: The "Cold Start" & The "Final Mile"

Every developer knows the loop. A new ticket lands. The first thirty minutes are a momentum-killing grind of context-switching, file-hunting, and boilerplate. Then, after hours of brilliant coding, you face the "final mile": another thirty minutes of writing commit messages, crafting PR descriptions, running quality checks, and updating tickets.

This friction at the beginning and end of every task is a massive drain on creativity and velocity.

The Solution: Your AI Apprentice

CodeScribe is a comprehensive agent ecosystem built to eliminate this friction entirely. We're not just another AI assistant; we are a proactive automation platform that handles the tedious work so developers can focus on what they do best: building.

  • 💡 The Idea Engine: For Product Managers, CodeScribe takes a high-level feature request and uses a Vellum workflow to generate a preliminary code diff, providing a tangible starting point for engineering.
  • 🚀 The Developer's Co-Pilot: For Developers, CodeScribe lives in the Warp terminal, providing a suite of commands to kickstart development and automate the entire PR submission process.
  • 🤖 The Team's Reviewer: For the whole team, CodeScribe acts as a tireless agent inside Linear, responding to @mentions to perform automated code reviews.

A Day in the Life with CodeScribe

This is not just a tool; it's a completely new way to work.

1. The Kickstart: From Ticket to Code in Seconds

A developer is assigned a ticket COD-42. Instead of starting from scratch, they run one command.

codescribe apply COD-42

Instantly, the AI-generated starter code from our Vellum workflow is applied to their local files. The "cold start" problem is solved.

2. The Finish Line: From Code to PR in One Command

After the developer has applied their expertise and finished coding, they run the ultimate command.

codescribe pr

This single command triggers a powerful automation cascade:

  • AI-Generated Commits: Analyzes changes and writes perfect, conventional commit messages.
  • Automated PRs: Creates or updates a GitHub pull request with a rich, AI-generated description.
  • Visual Documentation: For complex logic, it auto-generates Mermaid diagrams and embeds them in the PR.
  • Built-in Scans: Performs quick code quality and security checks.
  • Seamless Sync: Posts a final "Agent Action" back to the Linear ticket with the PR link.

The developer never leaves their terminal. The "final mile" is covered in seconds.


This experience taught us a critical lesson: a successful AI product isn't just about a clever prompt. It's about building a resilient, fault-tolerant system around the AI to manage its inherent unpredictability.

What's Next for CodeScribe

We're incredibly proud of what we built in a weekend, but we're just scratching the surface of what's possible. Our roadmap includes:

  1. IDE Extension: A "one-click apply" and "one-click PR" button directly in VS Code, bringing the agent's power even closer to the developer's workflow.
  2. Automated Test Generation: The agent will not only write the feature code but also generate corresponding unit test stubs to ensure code quality from the start.
  3. Self-Improving Feedback Loop: Allow developers to approve or reject the AI's changes. This feedback will be used to fine-tune the AI model, making it smarter and more aligned with a specific team's coding style over time.
  4. Interactive Patch Application: Instead of applying the entire diff at once, the agent will present individual "hunks" of code changes, allowing the developer to selectively accept or reject parts of the AI's suggestion.

Quick Start

Get up and running with CodeScribe in minutes.

1. Installation

Clone the repository and install dependencies.

git clone https://github.com/ManagementMO/CodeScribe.git
cd codescribe
npm install

# Set up your environment variable (BYO) for Gemini, Vellum, Linear, etc.
Example .env:
GITHUB_TOKEN=github_pat_...
LINEAR_AGENT_KEY=lagn_...
GEMINI_API_KEY=AIzaSy...
SUPABASE_URL=https://...supabase.co
SUPABASE_ANON_KEY=ey...

# Make the codescribe command globally available on your system.
npm link

# Basic Usage
Now you can use the CodeScribe agent from any git repository on your machine.

# Apply a preliminary diff from a Linear ticket
codescribe apply TICKET-ID

# Use it in warp! Press CTRL + R and search for the codescribe workflows to automate even further.

# Run the comprehensive PR creation/update workflow
codescribe pr

# See all available commands and options
codescribe --help 
---

About

An AI-powered agent ecosystem that automates the entire development pipeline, from generating initial code drafts to creating fully-packaged pull requests.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors