Nomad模型上下文协议服务器

go install github.com/danfedick/nomad-mcp@latest

git clone https://github.com/danfedick/nomad-mcp.git
cd nomad-mcp
go build -o nomad-mcp .

export NOMAD_ADDR="https://nomad.example.com:4646"
export NOMAD_TOKEN="your-acl-token"  # if ACLs enabled
export NOMAD_CACERT="/path/to/ca.pem"  # if using TLS

{
  "mcpServers": {
    "nomad": {
      "command": "/path/to/nomad-mcp",
      "env": {
        "NOMAD_ADDR": "https://nomad.example.com:4646",
        "NOMAD_TOKEN": "your-token"
      }
    }
  }
}

mcp:
  servers:
    nomad:
      command: /path/to/nomad-mcp
      env:
        NOMAD_ADDR: https://nomad.example.com:4646
        NOMAD_TOKEN: your-token

"List all jobs"
"Show me running jobs in the production namespace"
"What jobs are currently pending?"

"Get status of the 'web' job"
"Show me details for the api-server job in production"
"What's happening with the traefik job?"

"Deploy this job: [paste HCL spec]"
"Run this Nomad job and wait for it to be healthy"

"Stop the web job"
"Stop and purge the old-api job"
"Shut down the batch-processor in staging"

"Scale the web job to 5 instances"
"Scale down api-server to 2 replicas"
"Set the worker group in batch-job to 10"

"Show allocations for the web job"
"List all instances of the api service"

"Get status of allocation abc123"
"Show me details for alloc 7f3e"

"Show me logs for the web job"
"Get stderr from allocation abc123"
"Show the last 10000 bytes of logs from the api task"
"What errors is the worker showing?"

"Restart allocation abc123"
"Restart the nginx task in alloc 7f3e"
"Bounce the web service"

"List all nodes"
"Show me nodes that are ready"
"Which nodes are currently draining?"

"Show status of node abc123"
"What's running on node-1?"
"Get details for the node starting with 7f3e"

"Drain node-1 for maintenance"
"Stop draining node abc123"
"Drain node-2 with a 5 minute deadline"
"Enable drain on the node starting with 7f3e"

"List all deployments"
"Show recent deployments in production"
"What deployments are running?"

"Get status of deployment abc123"
"How's the latest deployment going?"

"Promote deployment abc123"
"The canaries look good, promote the deployment"
"Promote all task groups in the current deployment"

"List all services"
"What services are registered?"
"Show services in the production namespace"

"Get info for the web service"
"Show me the endpoints for api-server"
"Where is the redis service running?"

"List all variables"
"Show variables under nomad/jobs/"

"Get the variable at nomad/jobs/web/config"
"Read the database config variable"

"Store DATABASE_URL=postgres://... in nomad/jobs/api/config"
"Update the web config variable with new values"

"What's the cluster status?"
"Is the cluster healthy?"
"Who's the current leader?"
"Show me the server members"

"List recent evaluations"
"Show blocked evaluations"
"What evaluations are pending for the web job?"

You: "How's the web service doing?"

Agent: [calls nomad_job_status with job_id="web"]
       "The web service is running with 3 healthy allocations across 
        2 nodes. All instances are in the 'running' state. The last 
        deployment completed successfully 2 hours ago."

You: "The API seems slow, can you check the logs?"

Agent: [calls nomad_job_status with job_id="api"]
       [calls nomad_alloc_logs with the allocation ID]
       "I found some concerning entries in the API logs. There are 
        several timeout errors connecting to the database. The errors 
        started about 15 minutes ago. Would you like me to restart 
        the allocations or check the database service?"

You: "We're about to get a traffic spike, scale the frontend to 10"

Agent: [calls nomad_job_scale with job_id="frontend", group="web", count=10]
       "Scaled the frontend web group to 10 instances. The new 
        allocations are being placed now. I'll check back in a moment...
        [calls nomad_job_status]
        All 10 instances are now running and healthy."

You: "We need to patch node-3, can you drain it?"

Agent: [calls nomad_node_drain with node_id="node-3", enable=true]
       "I've enabled drain on node-3. Nomad is migrating 4 allocations 
        to other nodes. Current progress:
        - web: 2/2 migrated
        - api: 1/2 migrated (1 in progress)
        The node should be clear in about 2 minutes."

You: "Deploy this new version of the worker service"

Agent: [calls nomad_job_run with the spec]
       "Deployment started. Using canary strategy - 1 canary instance 
        is being launched first.
        [calls nomad_deployment_status]
        The canary is healthy and passing health checks. Ready to 
        promote when you give the go-ahead."

You: "Looks good, promote it"

Agent: [calls nomad_deployment_promote]
       "Promoted the deployment. Rolling out to remaining instances...
        All 5 instances updated and healthy. Deployment complete!"

namespace "*" {
  policy = "read"
}
node {
  policy = "read"
}

namespace "*" {
  policy = "write"
}
node {
  policy = "write"
}
agent {
  policy = "write"
}