CPLA MCP Server

Secure, Private AI Self-Awareness for GitHub Copilot

An MCP server that gives GitHub Copilot self-awareness about YOUR personal coding patterns - 100% local, zero cloud, completely private.

Part of the Agent-OS & Agent-Mesh ecosystem.

🔒 Security First

| Guarantee | Status | |-----------|--------| | Data stays local | ✅ Never leaves your machine | | No cloud | ✅ Works completely offline | | No telemetry | ✅ Zero analytics or tracking | | Read-only | ✅ Never modifies your data | | Minimal deps | ✅ Only @modelcontextprotocol/sdk |

Read our full Security Policy →

🎯 What Makes This Different

| Tool | What It Does | |------|--------------| | Copilot Memory | Remembers project/repo context | | MemoriPilot | Tracks project decisions & architecture | | ContextIQ | Indexes code & stores rules | | CPLA | Understands YOU - your patterns, schedule, trust level |

CPLA is the only tool that gives Copilot self-awareness about the human using it.

📦 Install

npm install -g @imransiddique/cpla

⚡ Quick Setup

For GitHub Copilot CLI

Add to ~/.copilot/config.json:

{
  "mcpServers": {
    "cpla": {
      "command": "cpla-server"
    }
  }
}

For VS Code Copilot Chat

Add to VS Code settings.json:

{
  "github.copilot.chat.mcpServers": {
    "cpla": {
      "command": "cpla-server"
    }
  }
}

Restart Copilot

After configuring, restart your Copilot CLI or VS Code to load the MCP server.

💬 Start Using!

You: "What have I been working on?"
Copilot: "In the last 7 days you've focused on agent-os (40 mentions)
         and octane (31 mentions). You're a night coder - peak at 9 PM!"

You: "Should I auto-run tests?"
Copilot: "Yes! You approve test runs 94% of the time. Safe to automate."

You: "What skills should I add?"
Copilot: "Based on your patterns, add these to copilot-instructions.md:
         - Run tests before every commit
         - Use conventional commit messages"

🛠️ Available Tools

| Tool | Description | When Copilot Uses It | |------|-------------|---------------------| | get_my_profile | Full personality profile | "Who am I?", "What's my coding style?" | | get_my_activity | Recent requests & projects | "What have I been doing?", "Recent activity" | | get_my_patterns | Tool usage, workflows | "What are my patterns?", "How do I work?" | | get_my_schedule | Coding hours, peak times | "When do I code?", "Am I a night coder?" | | get_my_approvals | Trust level, responses | "Do I trust you?", "My approval rate?" | | get_my_projects | Projects worked on | "What projects?", "My focus areas?" | | suggest_skills | Personal skill suggestions | "What should I automate?", "Improve my workflow" | | should_i_automate | Check if action is auto-worthy | "Should I auto-run tests?" |

📦 From Source (Development)

1. Install

cd mcp-server
npm install

2. Build

npm run build

3. Test

npm test

4. Configure Copilot

Add to your Copilot MCP settings (~/.copilot/config.json or VS Code settings):

{
  "mcpServers": {
    "cpla": {
      "command": "node",
      "args": ["path/to/cpla-mcp-server/dist/index.js"]
    }
  }
}

🧪 Testing

CPLA has comprehensive test coverage with 78 tests across 14 test suites:

| Area | Tests | What's Covered | |------|-------|----------------| | Constants | 6 | Categories, keywords, approval/denial words | | Activity Classification | 10 | All 8 categories, edge cases, case sensitivity | | Project Detection | 5 | Counting, sorting, all keywords | | Approval Analysis | 11 | Rates, thresholds, mixed scenarios | | Schedule Analysis | 11 | All coder types, time buckets, boundaries | | Tool Counting | 3 | Aggregation, empty input | | Phrase Extraction | 4 | N-gram extraction, frequency filtering | | Integration | 22 | All 8 MCP tools with fixture data | | Edge Cases | 6 | Missing dirs, old sessions, empty input |

Run tests with:

npm test

🔒 Privacy

• 100% Local: All data stays on your machine
• No Cloud: Nothing is sent anywhere
• Your Data: Reads only from ~/.copilot/session-state/
• Transparent: See exactly what it analyzes

📊 Data Sources

CPLA reads from your existing Copilot CLI session data:

~/.copilot/session-state/
├── [session-id]/
│   └── events.jsonl    # Your interactions

It analyzes:

• User messages (your requests)
• Tool executions (what Copilot did)
• Timestamps (when you code)
• Response patterns (approvals/denials)

🚀 Why This Matters

1. Personalized Assistance: Copilot can now tailor responses to YOUR style
2. Proactive Suggestions: "You usually run tests here, should I?"
3. Skill Discovery: Automatically suggests skills based on YOUR patterns
4. Trust Calibration: Knows when to ask vs. when to just do it

🎬 Example Session

You: "I want to improve my Copilot experience"

Copilot: Let me analyze your patterns...

[calls get_my_profile]

Based on your last 14 days:

🎯 You mainly do: Automation/Agents (197), PR/Code Review (180), Exploring (167)
📁 Your projects: agent-os, octane, agent-mesh
✅ Trust level: 90% approval rate - you trust me!
🕐 Schedule: Night coder (peak at 9 PM)

Suggestions for your copilot-instructions.md:
- "Run tests before every commit" (you approve 94% of test runs)
- "Use conventional commits" (you do 91 git operations)
- "Be careful with late-night changes" (you code until 1 AM)

Want me to add these to your instructions?

📝 License

UNLICENSED

🌐 Part of the Agent-OS Ecosystem

CPLA is built on the same security-first principles as:

• Agent-OS - Secure agent orchestration framework
• Agent-Mesh - Distributed agent communication

Together, these tools provide a complete, secure AI agent infrastructure.

"Secure AI that understands how YOU use AI"

Built by Imran Siddique with security and privacy at the core.