Skip to main content
sudocode is a lightweight, git-native context management system for AI-assisted software development. It helps you capture requirements, track agent work, and maintain bidirectional connections between specifications and implementation, all living directly in your repository. Think of sudocode as “context-as-code”: a structured way to organize collaboration between humans and AI agents throughout your development workflow.
  • Specs capture user intent and requirements
  • Issues track agent tasks and implementation work
  • Relationships model dependencies between work items
  • Feedback enables agents to improve specs as they implement
Everything is version-controlled alongside your code. No external services required. sudocode in action

​
Why sudocode?

Learn Why sudocode Matters

Understand the problems sudocode solves and how it transforms AI-assisted development

​
Who is sudocode for?

​
Perfect For

AI-Assisted Development Teams

Teams using Claude Code, Cursor, Copilot, or custom agents for daily development work

Long-Horizon Projects

Projects spanning weeks/months where context needs to persist across many sessions

Complex Requirements

Features with many interdependent tasks, dependencies, and design decisions to track

Multi-Agent Workflows

Running multiple AI agents concurrently (e.g., one on backend, one on frontend)

Spec-Driven Development

Teams that value written requirements and want iterative refinement from implementation

Open Source Projects

Need context that lives in the repo, accessible to all contributors without external accounts

​
Maybe Not For

  • Quick prototypes: Single-session exploratory coding without need for persistence
  • Solo scripts: One-off automation where there’s no ongoing maintenance
  • Non-AI workflows: Teams not using AI coding assistants (though sudocode still works as a lightweight issue tracker)

​
Quick Start

Get up and running in minutes:

Quick Start Guide

Create your first spec and issue in under 5 minutes

​
Learn the Core Concepts

Understand how sudocode organizes context:

Specs

Capture user requirements and design decisions

Issues

Track agent tasks and implementation work

Relationships

Model dependencies and connections between work

Feedback

Enable bidirectional learning between agents and specs

​
Integration Guides

Connect sudocode to your AI agents and workflows:

Quick Start

Set up sudocode and create your first issue in minutes

Workflows

Learn spec-driven development and best practices

MCP Integration

Use sudocode with Claude Code and other MCP-compatible agents

​
Key Features

  • Dual representation: Human-readable markdown + machine-queryable JSONL
  • Graph-based planning: Topological ordering finds ready work automatically
  • Bidirectional linking: Obsidian-style [[SPEC-001]] references with backlinks
  • Smart anchoring: Feedback tracks specific lines and relocates intelligently
  • Distributed sync: Git handles distribution; AI handles merge conflicts
  • Multiple interfaces: CLI, MCP server, REST API, and web UI

​
Community & Support

GitHub Discussions

Ask questions and share use cases

GitHub Issues

Report bugs and request features
Philosophy: Context should be treated like code—version-controlled, collaboratively edited, and evolved alongside your implementation. sudocode makes this possible for AI-assisted development.