Ask Claude "Why is production down?" and get a real answer from your live systems.
An MCP server that connects Claude to your DevOps stack. Query Kubernetes, Jenkins, GitHub Actions, AWS, Azure, GCP, ServiceNow, Slack, and Microsoft Teams in plain English without switching between tabs and terminals.
You: "The API deployment has been stuck for 20 minutes. What's going on?"
Claude: I'll check the deployment status and scan for failing pods.
[k8s_get_deployment_status("api-server", "production")]
-> desired: 3 ready: 1 updated: 3 strategy: RollingUpdate
-> Condition: Progressing - ReplicaSetUpdated
[k8s_get_failing_pods("production")]
-> api-server-7d9f8b-xk2p9 CrashLoopBackOff restarts: 14 node: node-1
-> api-server-4k8m3-rz9q1 CrashLoopBackOff restarts: 11 node: node-2
[k8s_pod_logs("api-server-7d9f8b-xk2p9", tail_lines=30)]
-> [09:10:03] ERROR postgresql://db-prod:5432: Connection refused
-> [09:10:03] FATAL Could not initialize connection pool. Exiting.
[k8s_get_events("production", event_type="Warning")]
-> OOMKilling Pod/db-prod-0 "Memory limit exceeded" count: 3 09:09:47
[snow_get_incident("INC0045231")]
-> short_description: API gateway 502 errors spiking
-> state: In Progress assigned_to: oncall-sre
Root cause: db-prod-0 was OOMKilled at 09:09:47. The two API pods are
crash-looping because the database stopped accepting connections.
INC0045231 is already open and assigned to oncall-sre.
Recommended next steps:
kubectl describe pod db-prod-0 -n production # check memory stats
kubectl logs db-prod-0 -n production # confirm OOM in db logs
Raise db-prod memory limit: 512Mi -> 1Gi
+---------------------------------------------------------------------------------+
| Claude (LLM) |
| "Why is production down?" / "Which incidents are open right now?" |
+----------------------------------+----------------------------------------------+
| MCP Protocol (stdio)
v
+---------------------------------------------------------------------------------+
| devops-mcp server |
| |
| +-------------+ +-------------+ +-------------+ +-------------+ |
| | Kubernetes | | Jenkins/GHA | | AWS | | Azure | |
| | pods/logs | | build status| | CW logs | | Monitor | |
| | events | | build logs | | metrics | | AKS status | |
| | rollouts | | failing jobs| | S3 / ECS | | res. health | |
| +-------------+ +-------------+ +-------------+ +-------------+ |
| |
| +-------------+ +-------------+ +-------------+ +-------------+ |
| | GCP | | ServiceNow | | Slack | | Teams | |
| | Cloud Log | | incidents | | post alerts | | post alerts | |
| | Monitoring | | changes | | history | | messages | |
| | GKE / Run | | | | search | | channels | |
| +-------------+ +-------------+ +-------------+ +-------------+ |
| |
| Audit log (~/.devops-mcp/audit.log) |
+---------------------------------------------------------------------------------+
| Tool |
What it answers |
k8s_get_failing_pods |
What pods are crashing right now? |
k8s_pod_logs |
What is this pod printing to stdout/stderr? |
k8s_describe_pod |
Full state: conditions, restart counts, recent events |
k8s_get_events |
What Warning events has the cluster emitted recently? |
k8s_get_deployment_status |
Is this rollout progressing or stuck? |
| Tool |
What it answers |
jenkins_get_build_status |
Did build #N succeed or fail, and how long did it take? |
jenkins_get_build_log |
What did the last N lines of the build print? |
jenkins_list_failing_jobs |
Which jobs are currently red? |
| Tool |
What it answers |
gh_get_workflow_run |
Status and conclusion for a specific run |
gh_list_failed_runs |
Which workflows have failed in the last 24 hours? |
| Tool |
What it answers |
aws_get_cloudwatch_logs |
What did this log group emit matching this filter? |
aws_get_metric |
What does this CloudWatch metric look like right now? |
aws_describe_s3_bucket |
Versioning, lifecycle rules, encryption config |
aws_get_ecs_service_status |
Running vs desired vs pending task counts |
| Tool |
What it answers |
azure_get_monitor_logs |
Run a KQL query against a Log Analytics workspace |
azure_get_metric |
What does this Azure Monitor metric look like right now? |
azure_get_aks_node_status |
AKS agent pool provisioning state and node counts |
azure_get_resource_health |
Is this Azure resource healthy? |
| Tool |
What it answers |
gcp_get_logs |
Query Cloud Logging with a filter expression |
gcp_get_metric |
What does this Cloud Monitoring metric look like? |
gcp_get_gke_cluster_status |
GKE cluster and node pool status |
gcp_get_cloud_run_status |
Cloud Run revision status and traffic split |
| Tool |
What it answers |
snow_list_incidents |
What incidents are open right now? |
snow_get_incident |
Full detail on a specific incident (INC0012345) |
snow_get_change_requests |
What changes are scheduled or in flight? |
| Tool |
What it does |
slack_post_message |
Post an alert or notification to a channel |
slack_get_channel_history |
What was said in #incidents in the last hour? |
slack_search_messages |
When was this topic last discussed across Slack? |
| Tool |
What it does |
teams_post_message |
Post an alert card to a channel via webhook |
teams_get_channel_messages |
What was said in this channel recently? |
teams_list_channels |
What channels exist in this team? |
git clone https://github.com/gerardrecinto/devops-mcp.git
cd devops-mcp
pip install -e .
# Optional extras for the integrations you use
pip install -e ".[azure]" # Azure Monitor + AKS
pip install -e ".[gcp]" # Cloud Logging + Monitoring + GKE + Cloud Run
pip install -e ".[slack]" # Slack SDK
# ServiceNow and Teams use requests, which is already included
cp .env.example .env
# Fill in credentials for each service. Any service without credentials is skipped.
Edit ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"devops-mcp": {
"command": "devops-mcp",
"env": {
"KUBECONFIG": "/Users/you/.kube/config",
"JENKINS_URL": "https://jenkins.example.com",
"JENKINS_USER": "your-user",
"JENKINS_TOKEN": "your-api-token",
"GITHUB_TOKEN": "ghp_xxxxxxxxxxxxxxxxxxxx",
"AWS_REGION": "us-west-2",
"AZURE_TENANT_ID": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
"AZURE_CLIENT_ID": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
"AZURE_CLIENT_SECRET": "your-client-secret",
"AZURE_SUBSCRIPTION_ID": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
"GCP_PROJECT_ID": "my-project-id",
"GOOGLE_APPLICATION_CREDENTIALS": "/Users/you/.gcp/service-account.json",
"SNOW_INSTANCE": "yourcompany.service-now.com",
"SNOW_USER": "your-user",
"SNOW_PASSWORD": "your-password",
"SLACK_BOT_TOKEN": "xoxb-xxxxxxxxxxxxxxxxxxxx",
"TEAMS_WEBHOOK_URL": "https://outlook.office.com/webhook/xxxx",
"TEAMS_TENANT_ID": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
"TEAMS_CLIENT_ID": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
"TEAMS_CLIENT_SECRET": "your-client-secret"
}
}
}
}Restart Claude Desktop. The devops-mcp tools appear in the toolbar.
cp .env.example .env # fill in credentials
docker-compose up -d
Infrastructure tools (Kubernetes, AWS, Azure, GCP, Jenkins, GitHub Actions) are read-only. Claude can observe and explain but cannot modify infrastructure. Slack and Teams tools can post messages for incident alerting.
| Action |
Available |
| Read pod logs, events, deployment status |
yes |
| Read build logs and pipeline status |
yes |
| Read CloudWatch / Azure Monitor / Cloud Logging |
yes |
| Read ServiceNow incidents and change requests |
yes |
| Post to Slack channels |
yes |
| Post to Teams channels via webhook |
yes |
kubectl apply / trigger builds / modify resources |
no |
| AWS / Azure / GCP resource creation or deletion |
no |
| Create or close ServiceNow records |
no |
Credentials are injected via environment variables, never hardcoded. Each integration is independently optional; missing credentials skip that tool group.
Audit log: every tool call is appended to ~/.devops-mcp/audit.log with ISO timestamp, tool name, and parameters.
pip install -e ".[dev]"
pytest tests/ -v --cov=devops_mcp --cov-report=term-missing
tests/test_azure.py::test_monitor_logs_demo_returns_events PASSED
tests/test_azure.py::test_get_metric_demo_returns_datapoints PASSED
tests/test_azure.py::test_aks_node_status_demo_returns_pools PASSED
tests/test_azure.py::test_resource_health_demo_returns_available PASSED
tests/test_gcp.py::test_get_logs_demo_returns_entries PASSED
tests/test_gcp.py::test_get_metric_demo_returns_points PASSED
tests/test_gcp.py::test_gke_status_demo_returns_running PASSED
tests/test_gcp.py::test_cloud_run_demo_returns_uri PASSED
tests/test_kubernetes.py::test_healthy_cluster_returns_message PASSED
tests/test_kubernetes.py::test_crashloop_pod_is_included PASSED
tests/test_kubernetes.py::test_pod_logs_returns_content PASSED
tests/test_kubernetes.py::test_describe_pod_is_valid_json PASSED
tests/test_kubernetes.py::test_demo_mode_on_import_error PASSED
tests/test_jenkins.py::test_successful_build_result PASSED
tests/test_jenkins.py::test_failed_build_result PASSED
tests/test_jenkins.py::test_list_failing_jobs_filters_correctly PASSED
tests/test_aws.py::test_cloudwatch_logs_demo_returns_events PASSED
tests/test_aws.py::test_get_metric_demo_returns_datapoints PASSED
tests/test_aws.py::test_ecs_service_status_demo_shows_degraded PASSED
tests/test_servicenow.py::test_list_incidents_demo_returns_incidents PASSED
tests/test_servicenow.py::test_get_incident_demo_returns_details PASSED
tests/test_servicenow.py::test_get_change_requests_demo_returns_changes PASSED
tests/test_slack.py::test_post_message_demo_returns_ok PASSED
tests/test_slack.py::test_get_channel_history_demo_returns_messages PASSED
tests/test_slack.py::test_search_messages_demo_returns_matches PASSED
tests/test_teams.py::test_post_message_demo_returns_ok PASSED
tests/test_teams.py::test_get_channel_messages_demo_returns_messages PASSED
tests/test_teams.py::test_list_channels_demo_returns_channels PASSED
28 passed in 0.52s
Coverage: 91%
PRs welcome. Open an issue first for significant new integrations.
MIT