cdp-pc — CDP Public Cloud control-plane MCP (read-only)

A Model Context Protocol server for monitoring the Cloudera CDP Public Cloud control plane in natural language, from any MCP client (Claude Code, Claude Desktop, …).

It fills a gap: existing Cloudera MCPs cover the data plane (Iceberg/Impala), streaming (SMM/SSB/NiFi), or private-cloud Cloudera Manager — none cover the CDP-PC control plane (environments / datalake / datahub / DW / DE / ML lifecycle).

v0.1 is read-only. Every tool only ever runs list- / describe- / get- verbs — enforced in code (cdp.py), so there is zero risk of mutating a customer environment.

How it works

Each tool shells out to the installed cdp CLI with --output json and returns the parsed result. This reuses your existing CLI install and ~/.cdp/credentials profiles — no separate auth, no cdpcli library coupling, no text scraping. Built on the official cloudera/minimcp framework.

Related MCP server: mcp-mgccli

Tools

Every tool also takes an optional profile (a profile name from ~/.cdp/credentials, default default).

Setup

Requires the cdp CLI installed and configured (~/.cdp/credentials).

cd cdp-pc-mcp
python3 -m venv .venv
. .venv/bin/activate
pip install -r requirements.txt

Register in Claude Code

BASE="$(pwd)"   # run this from the repo root
claude mcp add cdp-pc -s user -e "PYTHONPATH=$BASE/src" -- "$BASE/.venv/bin/python" -m cdp_pc_mcp.stdio_main

Then start a new Claude Code session and try: "list my CDP environments", "what datahubs do I have and their status", "describe environment my-cdp-env".

Quick self-test (no MCP client)

. .venv/bin/activate && export PYTHONPATH="$PWD/src"
python -c "from cdp_pc_mcp.cdp import run_cdp; print(run_cdp('iam','get-user')['user']['email'])"

Roadmap (v0.2+)

• Operations with guardrails: start/stop environment·datahub·ml, DW scaling

• HTTP transport for packaged customer delivery

• Richer diagnostics; packaging for contribution to the cloudera org