FlowSpec Plugin for Claude Code
A Claude Code plugin that turns FlowSpec JSON exports into your source of truth for spec-driven development. No API or MCP servers required — pure knowledge that works with any codebase.
What is FlowSpec?
FlowSpec is a visual data architecture tool. You annotate wireframes with data flows, business logic, and constraints on an infinite canvas, then export structured JSON specifications that describe every data element, UI component, transform, and the edges connecting them.
Skills
flowspec:spec — Spec-Driven Development
Load a FlowSpec JSON export and use it as a requirements document while building features. The spec defines your types, validation rules, data flows, and business logic — Claude references it as you work.
Use when:
- Generating TypeScript interfaces and Zod schemas from a spec
- Building pages/components that match the specified data architecture
- Implementing business logic transforms
- Creating API endpoints or database schemas from data points
- Any feature work where a FlowSpec export describes the requirements
Usage:
/flowspec:spec path/to/flowspec-export.json
flowspec:architect — MCP-Powered Data Architecture
Create and manage FlowSpec projects directly from Claude Code using the FlowSpec MCP server. Supports three workflows: wireframe-to-dataflow, codebase-to-dataflow, and design-new-screens. Requires the flowspec-mcp server configured.
Use when:
- Creating a new FlowSpec project from wireframe images or a codebase
- Building data architecture with nodes, edges, and screens via MCP
- Analysing and iterating on an existing FlowSpec project
Usage:
/flowspec:architect [wireframe images | codebase path | 'new project']
flowspec:architect-web — Browser-Based Data Architecture
Same three workflows as architect, but uses Chrome browser tools instead of MCP — for users without the desktop app. Generates JSON locally and uses light browser automation to import into the web editor at flowspec.app.
Use when:
- You don't have FlowSpec Desktop or the MCP server configured
- You want to work with the FlowSpec web app via Chrome automation
Usage:
/flowspec:architect-web [wireframe images | codebase path | 'new project']
flowspec:wireframe — Screenshot Component Design
Combine a wireframe screenshot with a sub-canvas JSON spec to implement UI components with both visual and data context. Claude sees the screenshot AND knows what data goes where via placement coordinates.
Use when:
- Implementing a specific UI section from a wireframe
- Building components where visual layout and data placement both matter
- Translating annotated wireframes into production code
Usage:
/flowspec:wireframe path/to/screenshot.png path/to/subcanvas.json
Optionally include the main spec for full type and constraint information:
/flowspec:wireframe path/to/screenshot.png path/to/subcanvas.json path/to/main-spec.json
Installation
From Marketplace (recommended)
If you have FlowSpec Desktop installed, the marketplace is auto-configured during setup:
claude plugin install flowspec@flowspec-plugins
Direct Installation
Without FlowSpec Desktop, install directly from GitHub:
claude plugin install flowspec@https://github.com/Jktfe/flowspec-claude-code-plugin
Note: Direct GitHub installation may not work due to SSH/authentication issues. Use the marketplace method or local install instead.
Local Development
Clone the repo and load it with --plugin-dir:
git clone https://github.com/Jktfe/flowspec-claude-code-plugin.git
claude --plugin-dir ./flowspec-claude-code-plugin
This loads the plugin for that session. To always load it, add the flag to your shell alias:
# In your .zshrc or .bashrc
alias claude='claude --plugin-dir /path/to/flowspec-claude-code-plugin'
How It Works
Both skills are pure knowledge — they teach Claude how to read and use FlowSpec JSON exports. No external services, MCP servers, or API keys are needed.
- Spec skill: Reads your JSON spec, maps DataPoints to types, Components to routes, Transforms to functions, and DataFlow edges to your dependency graph
- Wireframe skill: Reads your screenshot (multimodal vision) and sub-canvas JSON (percentage coordinates), then combines visual analysis with data placement to build accurate components
Example Workflow
- Design your data architecture in FlowSpec
- Export the main JSON spec and any wireframe sub-canvases
- Start a Claude Code session with the plugin loaded (see Installation above)
- Load the spec:
/flowspec:spec ./flowspec-export.json
- Ask Claude to build features: "Generate TypeScript types for the user profile section"
- For specific UI sections:
/flowspec:wireframe ./profile-header.png ./profile-header.json
License
