mcp-deploy-intel 0.5.0

Kubernetes deployment intelligence — typed query tools + AI-synthesised risk briefs. CLI + MCP server.

mcp-deploy-intel

Kubernetes deployment intelligence — typed query tools + AI-synthesised risk briefs. CLI + MCP server.

Status: v0.5.0 — 8 tools live + continuous watcher + outcome evaluator. Two K8s-only tools + three Prometheus tools + three SQLite deploy-history tools + K8s informer loop + async scoring loop (ROLLED_BACK / DEGRADED / SUCCESS / UNKNOWN). LLM risk-brief tools arrive in v0.6.0+; v1.0.0 adds Helm Deployment mode + cluster matrix CI.

Install

# Run ephemerally via uvx (recommended for one-off queries):
uvx mcp-deploy-intel --help

# Persistent:
pip install mcp-deploy-intel

Quickstart

# List Deployments + StatefulSets in a namespace
deploy-intel list-workloads-in-namespace default

# Snapshot a single workload
deploy-intel get-workload default/Deployment/my-api --format md

# Watch cluster deployments — auto-record every new rollout to SQLite
deploy-intel watch                               # all namespaces
deploy-intel watch --namespace prod --namespace staging   # scoped

# Score pending deploys — writes ROLLED_BACK / DEGRADED / SUCCESS / UNKNOWN
deploy-intel evaluate                            # defaults: window=600s, interval=60s
deploy-intel evaluate --window-s 300 --interval-s 30 --prom-url http://prometheus:9090

MCP — Claude Desktop / Claude Code

{
  "mcpServers": {
    "deploy-intel": {
      "command": "uvx",
      "args": ["mcp-deploy-intel", "serve-mcp"]
    }
  }
}

Current tools exposed via MCP (8 total): get_workload, list_workloads_in_namespace, get_metric_trend, get_inbound_traffic, get_outbound_traffic, record_deploy, get_deploy_history, get_rollback_history.

Design & plans

• Spec: docs/superpowers/specs/2026-04-21-mcp-deploy-intel-design.md
• Plan 1 (this release): docs/superpowers/plans/2026-04-21-mcp-deploy-intel-plan-1-walking-skeleton.md