breaking-brake/cc-wf-studio

What it solves

CC Workflow Studio provides a visual way to design AI agent orchestrations. It eliminates the need for "prompt-guessing" by allowing users to map out workflows on a canvas and export them into Markdown formats that various AI agents (such as Claude Code, GitHub Copilot, and Cursor) can natively understand and execute.

How it works

The system uses a central workflow.json file to define the logic. This file can be managed through three different interfaces:

1. VSCode Extension: A drag-and-drop React Flow canvas for visual design.
2. CLI (ccwf): A command-line tool for rendering, validating, exporting, and running workflows in environments like CI or SSH.
3. MCP Server: A Model Context Protocol (MCP) server that allows external AI agents to read and edit the workflow.json file directly using natural language.

Once designed, workflows are exported as agent skills or slash commands (e.g., .md files) into specific directories (like .claude/agents/ or .cursor/skills/) so the target AI agent can use them.

Who it’s for

Developers and AI engineers who want to visually architect multi-agent workflows, sub-agent orchestration, and MCP tool integrations without writing raw prompt files manually.

Highlights

• Visual Canvas: Drag-and-drop editor for designing agentic engineering patterns.
• Multi-Agent Support: Exports to formats compatible with Claude Code, GitHub Copilot Chat/CLI, OpenAI Codex, Zoo Code, Gemini CLI, Antigravity, and Cursor.
• AI-Assisted Editing: Use an MCP server to let AI agents iteratively refine the workflow canvas via natural language.
• Flexible Interfaces: Full parity between the VSCode extension, a CLI, and an MCP server.