CML Node Operations

MCP Server

• Command: cml-mcp (pip-installed, stdio transport)
• Requires: CML_URL, CML_USERNAME, CML_PASSWORD environment variables
• Optional: cml-mcp[pyats] for CLI execution via pyATS

Available Tools

Node Lifecycle

| Tool | Parameters | What It Does | |------|-----------|-------------| | start_node | lab_id/lab_title, node_id/node_label | Start (boot) a single node | | stop_node | lab_id/lab_title, node_id/node_label | Stop (shutdown) a single node | | get_node | lab_id/lab_title, node_id/node_label | Get node details: state, CPU, RAM, interfaces, config | | get_nodes | lab_id/lab_title | List all nodes with their current states | | wipe_node | lab_id/lab_title, node_id/node_label | Wipe a node's configuration (reset to default) |

Configuration Management

| Tool | Parameters | What It Does | |------|-----------|-------------| | set_node_config | lab_id/lab_title, node_id/node_label, config | Set a node's startup configuration | | get_node_config | lab_id/lab_title, node_id/node_label | Get a node's current configuration | | download_lab_configs | lab_id/lab_title | Download all node configs from a running lab |

Console & CLI Access

| Tool | Parameters | What It Does | |------|-----------|-------------| | get_node_console | lab_id/lab_title, node_id/node_label | Get console connection details (URL, protocol) | | get_node_console_log | lab_id/lab_title, node_id/node_label, lines? | Retrieve console log output from a running node | | execute_command | lab_id/lab_title, node_id/node_label, command | Execute a CLI command on a running node (requires pyATS) |

Node States

| State | Meaning | |-------|---------| | DEFINED_ON_CORE | Node exists but is not running | | BOOTING | Node is in the process of booting | | BOOTED | Node has finished booting and is reachable | | STOPPED | Node has been shut down |

Workflow: Configure and Boot a Node

When a user wants to configure a specific node:

1. Check node state: get_node to see current state
2. Set startup config: set_node_config with the full configuration
3. Start the node: start_node to boot it
4. Wait for boot: Nodes take 30-120 seconds depending on type
5. Verify: execute_command to run show version or show ip int brief
6. Report: Tell the user the node is ready

Workflow: Collect Running Configs

When a user wants to back up or review configs:

1. Ensure lab is running: get_lab to check state
2. Download all configs: download_lab_configs gets everything at once
3. Or target specific nodes: get_node_config for individual nodes
4. Optionally commit to GitHub: Use github-ops skill to commit configs to a repo

Workflow: Console Log Troubleshooting

When a user needs to see what happened during boot or troubleshoot a node:

1. Check node state: get_node — node must be BOOTED or BOOTING
2. Get console log: get_node_console_log to retrieve recent console output
3. Analyze output: Look for error messages, boot failures, config errors
4. Common issues to spot:

• %PARSER-4-BADCFG — invalid configuration commands
• %LINK-3-UPDOWN — interface state changes
• %OSPF-5-ADJCHG — OSPF adjacency changes
• %BGP-5-ADJCHANGE — BGP neighbor state changes
• %SYS-5-RESTART — system restart messages
• Boot failures, license errors, memory allocation errors

1. Report: Summarize the console log findings

Workflow: Execute CLI Commands

When a user wants to run show commands or make live changes:

1. Verify node is BOOTED: get_node to check state
2. Execute command: execute_command with the CLI command
3. Parse output: Present the results in a readable format
4. Multiple commands: Run them sequentially, one execute_command per command

Common commands to execute:

show ip interface brief
show ip ospf neighbor
show ip bgp summary
show running-config
show cdp neighbor
show version
show ip route
show interfaces status

Workflow: Reset and Reconfigure

When a user wants to start fresh with a node:

1. Stop the node: stop_node
2. Wipe the config: wipe_node resets to factory default
3. Set new config: set_node_config with the new startup configuration
4. Start the node: start_node
5. Verify: Execute show commands to confirm the new config took effect

Startup Configuration Templates

IOSv Router (Basic OSPF)

hostname {HOSTNAME}
!
interface Loopback0
 ip address {LOOPBACK_IP} 255.255.255.255
!
interface GigabitEthernet0/0
 ip address {MGMT_IP} {MASK}
 no shutdown
!
interface GigabitEthernet0/1
 ip address {P2P_IP} {MASK}
 no shutdown
!
router ospf 1
 router-id {LOOPBACK_IP}
 network {LOOPBACK_IP} 0.0.0.0 area 0
 network {P2P_NETWORK} {WILDCARD} area 0
!
line con 0
 logging synchronous
line vty 0 4
 login local
 transport input ssh
!
end

IOSv Router (Basic BGP)

hostname {HOSTNAME}
!
interface Loopback0
 ip address {LOOPBACK_IP} 255.255.255.255
!
interface GigabitEthernet0/1
 ip address {PEER_IP} {MASK}
 no shutdown
!
router bgp {LOCAL_AS}
 bgp router-id {LOOPBACK_IP}
 bgp log-neighbor-changes
 neighbor {NEIGHBOR_IP} remote-as {REMOTE_AS}
 !
 address-family ipv4
  network {LOOPBACK_IP} mask 255.255.255.255
  neighbor {NEIGHBOR_IP} activate
 exit-address-family
!
end

NX-OS Switch (Basic VXLAN Leaf)

hostname {HOSTNAME}
!
feature ospf
feature bgp
feature nv overlay
feature vn-segment-vlan-based
!
interface loopback0
  ip address {LOOPBACK_IP}/32
!
interface Ethernet1/1
  description To-Spine1
  no switchport
  ip address {P2P_IP}/30
  no shutdown
!
router ospf 1
  router-id {LOOPBACK_IP}
!
router bgp {AS}
  router-id {LOOPBACK_IP}
  neighbor {SPINE_IP} remote-as {AS}
    update-source loopback0
    address-family l2vpn evpn
      send-community extended
!
end

IOS-XR Router (Basic IS-IS)

hostname {HOSTNAME}
!
interface Loopback0
 ipv4 address {LOOPBACK_IP} 255.255.255.255
!
interface GigabitEthernet0/0/0/0
 ipv4 address {P2P_IP} {MASK}
 no shutdown
!
router isis CORE
 is-type level-2-only
 net 49.0001.{SYSTEM_ID}.00
 address-family ipv4 unicast
  metric-style wide
 !
 interface Loopback0
  passive
  address-family ipv4 unicast
  !
 !
 interface GigabitEthernet0/0/0/0
  address-family ipv4 unicast
  !
 !
!
end

Important Rules

• Check node state before operations — don't start an already-booted node or stop an already-stopped one
• Wait for boot — IOSv takes ~60s, NX-OS ~90s, IOS-XR ~120s to boot
• execute_command requires pyATS — the cml-mcp[pyats] extra must be installed
• Set configs BEFORE starting — set_node_config sets the startup config; apply it before boot
• Use labels not IDs — labels like "R1" are more human-readable than UUIDs
• Record in GAIT — log all configuration changes and CLI commands for audit trail

Environment Variables

• CML_URL — CML server URL
• CML_USERNAME — CML username
• CML_PASSWORD — CML password
• CML_VERIFY_SSL — Verify SSL certificate (true/false)