specfront-ai 0.3.3

Spec-Driven AI Agent Architecture with MCP Integration

Specfront AI - Spec-Driven Copilot Agents

Specfront AI provides a "zero-friction" architecture for Spec-Driven Development (SDD) natively inside your GitHub Copilot workspace. By using .github/agents/*.agent.md files, this package instantly configures a Supervisor Agent and specialized Worker Agents (Planner, Coder, Tester) in your IDE, without needing complex LangChain orchestrators or external API keys.

It also includes an integrated FastMCP Server to pull context directly from Jira and Figma into your Copilot Chat.

🌟 Features

• Native Copilot Integration: Agents live right in your IDE chat. Just type @specfront_ai.
• Multi-Agent Architecture: Delegate tasks to specialized personas:
  • @specfront_ai (Supervisor): Orchestrates the workflow and delegates commands.
  • @specfront_ai_planner: Focuses strictly on writing Markdown Specifications and Jira tickets.
  • @specfront_ai_coder: Implements code based only on the strict requirements of the spec.
  • @specfront_ai_tester: Writes test suites to prove the spec's Acceptance Criteria are met.
• Official MCP Integration: Automatically generates an mcp-servers.json configuration file, securely linking your IDE to the official GitHub, Figma, and Jira Model Context Protocol servers.

---

🚀 Installation & Usage

Because Specfront AI is packaged purely as a CLI tool, installation takes seconds.

1. Install via PyPI

You can install the package directly into your virtual environment from the Python Package Index once published:

pip install specfront_ai

2. Scaffold the Agents

Open your existing application project in VS Code / Cursor. In your terminal, run:

specfront_ai init

What happens? The CLI will instantly generate:

1. The .github/agents/ prompt files.
2. A specs/ directory with a sample Hotel Management specification.
3. An mcp-servers.json file containing the precise initialization commands for the official GitHub, Jira, and Figma servers.

Copilot will automatically detect these .agent.md files—no restarts needed.

3. Talk to the Agent

Open Copilot Chat and invoke the supervisor:

@specfront_ai Hello! I want to start building the Hotel Inventory module.

Or invoke a specialized agent directly to get to work:

@specfront_ai_coder /implement Please read the hotel_inventory.md spec and create the Django models.

---

🔌 Activating the Official MCP Servers

To give your Copilot agents extra superpowers (like reading Jira tickets or inspecting Figma design tokens), you need to provide your exact API keys to the generated mcp-servers.json.

1. Open mcp-servers.json in your workspace root.
2. Replace the placeholder <YOUR_GITHUB_TOKEN>, <YOUR_JIRA_TOKEN>, and <YOUR_FIGMA_TOKEN> with your actual security tokens.
3. Point your IDE (Cursor, VS Code via Claude Desktop, etc.) to read this configuration file. The IDE will automatically install the servers via npx and attach them to your Copilot chats.

---

🛠️ Development & Publishing

If you want to contribute to the CLI or publish it to the official PyPI registry:

1. Read the pyproject.toml file to see the build configuration.
2. Consult PIP_PACKAGING_GUIDE.md for exact build and twine publishing commands.
3. Daily Updates: When making everyday updates, remember to increment the version in pyproject.toml before running python -m build and twine upload dist/*. See the packaging guide for the full redeployment workflow.