What AI coding agents does Draft support?

Draft supports Claude Code, Cursor, Codex, opencode, GitHub Copilot, and Google Gemini. It integrates natively with each platform using their respective configuration formats.

What is Draft?

Q: Does Draft work with existing projects?

Yes. Running /draft:init performs a 5-phase analysis of your existing codebase, generating architecture documentation, AI context files, and signal classifications. It works with any language, framework, or project structure — brownfield or greenfield.

Q: Does Draft require configuration?

Minimal. After installation, run /draft:init and it automatically analyzes your codebase and generates all necessary configuration files (product.md, tech-stack.md, architecture.md, etc.). Defaults work out of the box.

Context-Driven Development plugin for AI coding agents

Draft is a free, open-source Context-Driven Development (CDD) plugin that adds structured methodology to AI coding agents. It works with Claude Code, Cursor, Codex, opencode, GitHub Copilot, and Google Gemini to enforce spec-driven development, test-driven development (TDD), and automated code review. Draft provides 33 commands and 7 specialized agents that turn AI assistants from autonomous guessing machines into disciplined executors of pre-approved work.

The core idea: before AI writes a single line of code, it analyzes your codebase, generates a specification, builds a phased plan, and waits for your approval. Only then does implementation begin — constrained by your architecture, your conventions, and your acceptance criteria. Every decision is grounded in explicit, versioned, reviewable documents rather than implicit assumptions buried in a chat window.

How Draft Works

Draft operates through a graph-first, phase-driven workflow that mirrors how experienced engineering teams ship software:

Phase 0

Graph

/draft:init

Builds a knowledge graph of your codebase — callers, exports, hotspots, blast radius — powered by codebase-memory-mcp (159 languages, 100% local).

Phase 1

Init

/draft:init

5-phase codebase analysis on top of the graph. Generates architecture docs, AI context files, and project configuration.

Phase 2

Plan

/draft:plan

Creates a specification and phased implementation plan through collaborative dialogue. You approve before code begins.

Phase 3

Implement

/draft:implement

Builds features task-by-task with TDD cycles (RED → GREEN → REFACTOR) and architecture checkpoints.

Phase 4

Review

/draft:review

3-stage code review: automated validation, spec compliance, and code quality assessment.

This workflow is supported by 24 specialist commands for debugging, operations, quality assurance, documentation, and more. Each command loads only the context it needs, keeping token usage efficient while maintaining full project awareness.

Key Features

Canonical commands organized as 4 primary workflow commands + 5 routers dispatching to 24 specialist commands. Prefer the routers (`/draft:plan`, `/draft:ops`, `/draft:docs`, `/draft:discover`, `/draft:jira`) for most use.
7 specialized agents — Architect, Debugger, Planner, Reviewer, RCA (root cause analysis), Ops, and Writer — each with behavioral protocols tuned for their domain
Context tiering — three-layer system (profile → context → architecture) that loads the right amount of project knowledge for each task
Tier-gated init output — /draft:init packages architecture context as a single graph-primary architecture.md for small repos, or an OKF concept taxonomy under draft/wiki/ (one concept per file, cross-links form the graph) for larger ones, with .ai-context.md as the index root either way (DRAFT_INIT_MODE overrides)
TDD enforcement — RED → GREEN → REFACTOR cycle with test-first implementation and architecture checkpoints
3-stage code review pipeline — automated validation, spec compliance verification, and code quality assessment
ACID compliance auditing — enterprise-grade module lifecycle and data integrity verification
Pattern learning and guardrails — auto-learns your codebase conventions and enforces them; supports pattern promotion and false-positive elimination
Monorepo federation — discovers services, builds dependency graphs, and aggregates context across multiple packages
14-dimension bug hunting — exhaustive analysis with severity-ranked reports and regression test generation
Jira integration — maps tracks to epics, phases to stories, and tasks to sub-tasks with auto-generated story points
Incremental refresh — file-level freshness tracking with signal classification across 11 categories
Architecture Decision Records — documents significant technical decisions with context, alternatives, and consequences

Supported Platforms

Draft integrates natively with each platform using their respective configuration formats:

Platform	Integration Method	Setup
Claude Code	Native plugin (`npx @drafthq/draft install claude-code` or `/plugin install`)	30 seconds
Cursor	Native `.claude-plugin/` structure	30 seconds
Codex	`AGENTS.md` (read automatically)	30 seconds
opencode	`AGENTS.md` + global `skills/` directory	30 seconds
GitHub Copilot	`.github/copilot-instructions.md`	1 minute
Google Gemini	`.gemini.md` bootstrap file (incl. Antigravity IDE)	1 minute

Installation

For Claude Code (the primary platform):

# Install Draft (recommended)
npx @drafthq/draft install claude-code

# Or via plugin marketplace (alternative)
/plugin marketplace add drafthq/draft
/plugin install draft

# Initialize your project
/draft:init

For other platforms, see the installation guide on GitHub.

Frequently Asked Questions

Is Draft free?

Yes. Draft is completely free and open source under the MIT license. There are no paid tiers, no usage limits, and no vendor lock-in. The only costs come from the host AI tools you already use (Claude Code, Copilot, etc.).

How is Draft different from other AI coding tools?

Draft is not an AI coding tool — it is a methodology layer that runs on top of existing AI agents. While tools like Cursor or Copilot generate code, Draft ensures that code follows approved specifications, passes TDD gates, and fits your architecture. It adds structure and quality gates, not another AI model. There is no API key, no hosted service, no vendor lock-in. Draft is markdown files and bash scripts.

Does Draft work with existing projects?

Yes. Running /draft:init performs a 5-phase analysis of your existing codebase, generating architecture documentation, AI context files, and signal classifications. It works with any language, framework, or project structure — brownfield or greenfield.

What is Context-Driven Development?

Context-Driven Development (CDD) is a methodology where AI coding agents operate from persistent, file-based project context rather than ephemeral chat prompts. Draft analyzes your codebase to generate architecture docs, then enforces a spec-first workflow: specifications and plans are written and reviewed before any code is generated. Every decision is explicit, versioned, and reviewable.

Does Draft require a lot of configuration?

Minimal. After installation, run /draft:init and it automatically analyzes your codebase and generates all necessary configuration files (product.md, tech-stack.md, architecture.md, etc.). Defaults work out of the box. You can customize afterward.

Can Draft work with monorepos?

Yes. Run /draft:init at each module root — it is scope-aware and links each module graph to the root graph for full cross-module understanding.

Learn More

Draft is built and maintained by Mayur Pise and is available on GitHub under the MIT license.

Read the Book GitHub Getting Started Guide