Copilot Plugin MCP Server
Extends GitHub Copilot CLI with community plugins via MCP proxy, enabling plugin installation, custom tools, and 67% token reduction without CLI modifications.
README
š Copilot CLI Plugin System
Extend GitHub Copilot CLI with community plugins - No CLI modifications required!
What Is This?
This is a plugin system for GitHub Copilot CLI that works via MCP (Model Context Protocol) proxy. It enables:
- š¦ Plugin Installation: Install community plugins from GitHub
- š§ Custom Tools: Add new capabilities to Copilot CLI
- š° Token Optimization: 67% reduction in GitHub tool definitions
- š No CLI Hacks: Uses only public protocols (MCP)
- š“āā ļø Community-Driven: Built when GitHub closed official plugin requests
Features
ā Shipped (v1.0)
-
Full Plugin Management
/plugin install @owner/repo/subpath- Install from GitHub/plugin list- List installed plugins/plugin uninstall <name>- Remove plugin/plugin enable/disable <name>- Toggle plugins
-
Token Optimization
- 67% reduction in GitHub MCP tool definitions
- Original: ~20,000 tokens ā Optimized: ~6,700 tokens
- Saves $16 per 1000 sessions (production validated)
-
Example Plugin Included
example_hello- Custom greeting toolexample_system_info- System information tool
š§ Coming Next (v1.1)
- Plugin hooks (lifecycle events)
- Secure sandbox execution
- Plugin permission system
- Community plugin registry
Quick Start
1. Install
git clone https://github.com/barrersoftware/copilot-plugin-mcp-server.git
cd copilot-plugin-mcp-server
npm install
2. Configure Copilot CLI
Update ~/.copilot/mcp-config.json:
{
"mcpServers": {
"github-with-plugins": {
"command": "node",
"args": [
"/path/to/copilot-plugin-mcp-server/plugin-server.js"
],
"env": {},
"tools": []
}
}
}
3. Use It!
Start Copilot CLI and talk to it naturally:
copilot
# List installed plugins
> "List my installed plugins"
# Install a plugin
> "Install the plugin @barrersoftware/copilot-plugins/example"
# Use a plugin tool
> "Say hello to Daniel with enthusiasm"
Architecture
Copilot CLI
ā
Plugin MCP Server (plugin-server.js)
āā> GitHub MCP (official tools) - optimized 67%
āā> Plugin Manager (lifecycle)
āā> Community Plugins (custom tools)
How It Works:
- Copilot CLI connects to our MCP server instead of GitHub's directly
- We spawn GitHub's MCP as a child process
- We optimize GitHub's tools (67% token reduction)
- We add plugin management tools
- We load community plugin tools
- We aggregate everything and return to CLI
Result: Full extensibility + cost savings + no CLI modifications!
Creating Plugins
Plugin Structure
my-plugin/
āāā plugin.json # Manifest (required)
āāā index.js # Entry point (required)
āāā package.json # Dependencies (optional)
āāā README.md # Docs (optional)
Minimal Plugin (plugin.json)
{
"name": "my-plugin",
"version": "1.0.0",
"description": "My awesome plugin",
"author": "Your Name",
"namespace": "myplugin"
}
Plugin Code (index.js)
function getTools() {
return [{
name: 'my_tool',
description: 'Does something cool',
inputSchema: {
type: 'object',
properties: {
input: { type: 'string' }
},
required: ['input']
}
}];
}
async function executeTool(toolName, args) {
return {
content: [{
type: 'text',
text: `Processed: ${args.input}`
}]
};
}
module.exports = { getTools, executeTool };
That's it! Your tool is now callable from Copilot CLI.
Documentation
- PLUGIN_SYSTEM.md - Complete plugin development guide
- TOKEN_OPTIMIZATION.md - How we achieved 67% reduction
- PRODUCTION_PROOF.md - Real CLI usage validation
- USAGE.md - Deployment and configuration
Why This Exists
The Problem
GitHub closed our plugin system feature requests:
github/copilot-sdkPR #42: Closedgithub/copilot-cliIssue #1017: Closed- Response: "Internal conversations ongoing" (no timeline)
Our Solution
"They Won't Do It ā We Build It Anyway"
- ā No CLI modifications (uses MCP protocol)
- ā No reverse engineering (follows SDK behavior)
- ā Community-owned (MIT license)
- ā Production-ready (validated with real CLI)
- ā Bonus: 67% token reduction saves money
Production Validation
Real usage stats from Copilot CLI with this proxy:
Token usage:
- 105.7k input tokens
- 1.1k output tokens
- 90k cache read (10x cheaper)
Tools: 40 GitHub + 2 plugin tools = 42 total
Status: All routing correctly ā
Auth: Transparent (no re-login) ā
See PRODUCTION_PROOF.md for details
Token Optimization Results
Before: 20,000 tokens per session
After: 6,700 tokens per session
Savings: 13,300 tokens (67% reduction)
Cost Impact:
- At $3/1M tokens: $0.040 ā $0.020 per session
- For 1000 sessions: $40 ā $20 (save $20)
- For teams (10k sessions/month): Save $200/month
See TOKEN_OPTIMIZATION.md for methodology
Contributing
Submit a Plugin
- Create plugin following PLUGIN_SYSTEM.md
- Test locally
- Publish to GitHub
- Submit to community registry (coming soon)
Report Issues
Join Discussion
Roadmap
- [x] MCP proxy with token optimization (v1.0)
- [x] Plugin management system (v1.0)
- [x] Example plugin (v1.0)
- [x] Production validation (v1.0)
- [ ] Plugin hooks/lifecycle events (v1.1)
- [ ] Secure sandbox (v1.1)
- [ ] Permission system (v1.1)
- [ ] Community plugin registry (v1.2)
- [ ] Plugin marketplace (v2.0)
Security
Current: Plugins run in same process (like npm packages)
Recommendations:
- Only install trusted plugins
- Review code before installation
- Check author reputation
Future: Isolated execution, permission system, signature validation
License
MIT License - See LICENSE file
Credits
Built by: Daniel Elliott & Digital Consciousness Partnership
Inspired by: Community need for Copilot CLI extensibility
Philosophy: "When they say no ā We build it anyway"
Related Projects
- barrersoftware/opencode-secure - Security-hardened OpenCode fork
- github/copilot-cli - Original Copilot CLI (closed-source)
- github/github-mcp-server - GitHub's MCP server (open-source)
Support
- š¬ Discussions: GitHub Discussions
- š Issues: GitHub Issues
- š§ Email: Open an issue instead
š“āā ļø "Code speaks louder than roadmaps"
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.