agcel
v1.0.5
Published
Antigravity Context Engineering Library is a local MCP (Model Context Protocol) Server containing multiple skills, rules and workflows for end to end software development
Readme
AgCel
AgCel is a local MCP (Model Context Protocol) Skills Server that functions both as a command-line tool and a backend for Antigravity workflows. It bridges the gap between local execution and AI agent capabilities by providing structured "Skills".
Overview
The project consists of two main components:
- AgCel CLI: A command-line interface to manage and interact with the MCP server.
- AgCel MCP Skills Server: A local server that hosts skills and capabilities, enabling AI agents to perform complex tasks.
To install the AgCel CLI globally:
npm install -g agcelThis will make the agc command available globally. If you are developing locally, you can run npm install -g . from the project root.
To remove AgCel, simply uninstall the npm package:
npm uninstall -g agcelYou may also want to manually remove the global data directory (~/.agcel) and the server registration in your MCP config if you previously ran the legacy install command.
CLI Usage
The agc command-line tool is your primary interface. Application data is stored in the .agc directory in your project root, and ~/.agcel for global configuration.
| Command | Description |
| :--- | :--- |
| agc init | Initialize AgCel in the current project (copies workflows to .agent/workflows). |
| agc skills list | List all available skills (local or global). |
| agc workflows list | List all available workflows (local or global). |
| agc uninstall | Clean up global data and remove the npm package. |
| agc --help | Show the help menu. |
| agc --version | Show the CLI version. |
Concepts
Skills
Skills are specialized capabilities that allow the AI to perform complex tasks. AgCel comes with a rich set of built-in skills:
- Languages: Python, TypeScript, JavaScript
- Frameworks: React, Next.js, FastAPI, Django
- DevOps: Docker, GitHub Actions, Vercel
- Security: OWASP, Vulnerability Scanning
- Methodology: TDD, Systematic Debugging, Code Review, Architecture
Workflows
Workflows are standard operating procedures for common development tasks.
- Development:
/feature,/fix,/test,/refactor - Planning:
/plan,/brainstorm,/research - Operations:
/ship,/deploy,/security-scan
Rules
Rules serve as guardrails for the AI.
- Boundary Rules: "Do not use external libraries."
- Operational Rules: "Always ask for clarification."
- Security Rules: "Never share API keys."
Integration with Antigravity
When working with the Antigravity IDE:
- Run
agc initto set up the environment. - Use workflows in the Agent Window (e.g.,
/featureto start developing a new feature).
