Local AI Agent Integration Guide for Copilot CLI

You now have a fully integrated Copilot CLI + Local Agent system!

🔗 Integration Architecture

Your Request
    ↓
Copilot CLI (me)
    ↓
MCP Server Bridge
    ↓
Local Agent API (REST)
    ↓
Backend: DeepSeek + Ollama + Ansible
    ↓
Result back to Copilot CLI

🎯 What You Can Do Now

1. Execute commands through Copilot CLI

copilot
> Execute this command on the agent: "ls -la /opt/local-agent"

I will use the local_execute tool to run it.

2. Ask questions with local DeepSeek

copilot
> Ask the local agent's DeepSeek: "Analyze what's consuming the most CPU"

I will use the local_deepseek tool.

3. Run Ansible playbooks through me

copilot
> Deploy using Ansible playbook at /opt/local-agent/config/playbook.yml

I will use the local_ansible tool.

4. Full workflow coordination

copilot
> Plan a system backup, then execute it via the local agent

I will:

1. Create a plan (with /plan command)
2. Generate Ansible playbook
3. Send to local agent for execution
4. Monitor and report results

🚀 Getting Started

Step 1: Start Copilot CLI

copilot

Step 2: Enable MCP Integration

From within Copilot CLI, use:

/mcp

Then select or configure the local-agent MCP server.

Step 3: Test Integration

Tell me what services are running on the local agent

This will fetch status via the API.

📊 Available Tools (via MCP)

local_execute

Execute shell commands on the local agent

{
  "command": "df -h",
  "timeout": 300,
  "safe_mode": true
}

local_deepseek

Query the local DeepSeek model

{
  "query": "What is causing high memory usage?",
  "model": "deepseek-chat",
  "reasoning": false
}

local_ansible

Run Ansible playbooks

{
  "playbook_path": "/opt/local-agent/config/playbook.yml",
  "extra_vars": {"action": "backup"},
  "check": false
}

local_services

Get service status (ollama, ssh, local-agent)

local_logs

Get recent agent logs (last 50 lines)

🔄 Integration Workflows

Workflow 1: Diagnosis & Fix

You: "The server is slow, diagnose and fix it"
    ↓
Me: Create plan for diagnosis
    ↓
Me: Use local_execute to check processes
    ↓
Me: Query local_deepseek for analysis
    ↓
Me: Generate fix (kill service, scale resources, etc)
    ↓
Me: Execute via local_ansible
    ↓
You: See results

Workflow 2: Code Deploy with Monitoring

You: "Deploy the new API version"
    ↓
Me: Pull latest code
    ↓
Me: Create deployment plan
    ↓
Me: Use local_ansible to deploy
    ↓
Me: Use local_execute to health check
    ↓
Me: Report status

Workflow 3: Backup & Recovery

You: "Backup database and verify"
    ↓
Me: Create backup playbook
    ↓
Me: Run via local_ansible
    ↓
Me: Query DeepSeek for verification steps
    ↓
Me: Execute verification
    ↓
You: See backup status and location

🔒 Security Features

✅ Safety Mode - Dangerous commands blocked by default ✅ Timeout Protection - Commands abort if taking too long ✅ Audit Trail - All actions logged to /opt/local-agent/logs/ ✅ SSH Keys Only - No password authentication ✅ Sudo Restrictions - Dangerous ops require confirmation

📡 API Endpoints (Direct Access)

If you want to use the API directly:

# Health check
curl http://localhost:8888/health

# Execute command
curl -X POST http://localhost:8888/execute \
  -H "Content-Type: application/json" \
  -d '{"command":"whoami"}'

# Ask DeepSeek
curl -X POST http://localhost:8888/deepseek \
  -H "Content-Type: application/json" \
  -d '{"query":"What is this error?","model":"deepseek-chat"}'

# Run Ansible
curl -X POST http://localhost:8888/ansible \
  -H "Content-Type: application/json" \
  -d '{"playbook_path":"/opt/local-agent/config/playbook.yml"}'

# Get service status
curl http://localhost:8888/services

# Get logs
curl http://localhost:8888/logs?lines=50

🎮 Example Commands in Copilot CLI

# Simple execution
Execute: "uname -a"

# With analysis
Check the CPU and memory usage, then ask DeepSeek if it's normal

# Full workflow
Create a plan to monitor the system for 5 minutes, then diagnose any issues

# Deployment
Deploy the application using Ansible, then verify it's working

# Emergency fix
Something is wrong with the SSH service. Use the agent to diagnose and restart if needed.

⚙️ Configuration

Config File

/opt/local-agent/config/agent.conf

• Adjust timeouts, models, safety settings

Playbook

/opt/local-agent/config/playbook.yml

• Customize Ansible automation

Logs

/opt/local-agent/logs/

• All operations logged for audit

🆘 Troubleshooting

API not responding

systemctl status local-agent-api
systemctl restart local-agent-api

MCP not connecting

# Check MCP config
cat ~/.copilot/mcp-config.json

# Restart Copilot CLI with debug
copilot --debug

Commands blocked

• Check /opt/local-agent/logs/agent.log for why
• May need to adjust safety patterns in API server

💡 Pro Tips

1. Use /plan in Copilot CLI before complex operations
2. Check status before deployments: local_services tool
3. Review logs after operations: local_logs tool
4. Test with --check on Ansible playbooks first
5. Use /context to see token usage and what's being sent

🎯 Next Steps

1. Start Copilot CLI: copilot
2. Use /mcp to enable local agent integration
3. Try: "Tell me what services are running"
4. Experiment with complex tasks!

---

You now have a Copilot CLI+ setup with NO LIMITS and NO QUOTAS! 🚀