Provides tools for managing Kubernetes resources (clusters, workloads, etc.) within the GreenNode (VNG Cloud) infrastructure, enabling AI assistants to perform cluster operations and resource management.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@GreenNode MCP Server list all virtual servers in project pro-xxxx"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
A monorepo of MCP (Model Context Protocol) servers for GreenNode products,
organized as a uv workspace.
Each product is an independent project under src/, sharing the greennode
Python namespace.
Project | Package | Description |
| Shared core (config/profile loading, IAM auth, HTTP client with retry/401, validators, discovery cache) — product servers import it instead of copying plumbing. | |
| 41 tools + 3 prompts for managing VKS (GreenNode Kubernetes Service): clusters, node groups, resource discovery (name→ID), quota, on-demand creation guides, and in-cluster Kubernetes resources. EKS-style |
More servers will be added as sibling projects under src/.
Related MCP server: VergeOS MCP Server
greennode-mcp/
├── pyproject.toml # uv workspace root (members = ["src/*"])
├── src/
│ ├── mcp-core/ # shared core (greennode.mcp_core)
│ └── vks-mcp-server/ # one product = one independent project
│ ├── pyproject.toml
│ ├── greennode/
│ │ └── vks_mcp_server/
│ └── tests/
└── docs/# Install all workspace members into a shared environment
uv sync
# Run a specific server
uv run vks-mcp-serveruv run python scripts/new_server.py <product> # e.g. vdbScaffolds src/<product>-mcp-server/ from templates/new-server (working
example tool + tests + Dockerfile, conventions baked in), registers it with
release-please, and prints the remaining steps. CI (lint/test/build) discovers
the new package automatically. Agent guidance lives in .claude/skills/
(new-mcp-server, release-mcp) and the tiered CLAUDE.md files (repo root =
monorepo conventions; each package has its own).
See each project's own README for configuration and usage.
Note: Create/update tools use typed Pydantic request bodies (structured DTOs) for better code-mode support and self-documenting schemas.
--auth-debug (temporary, opt-in)--auth-debug (env GRN_MCP_AUTH_DEBUG=1) makes the HTTP transport log a
redacted summary of every inbound request and expose an unauthenticated
GET /whoami that echoes the same summary. It is meant for measuring what an
upstream (e.g. the MCP Gateway) actually sends — token scheme, JWT header
(alg/kid), allow-listed claims (iss/aud/sub/scope/...), and any
X-GreenNode-* / X-Forwarded-* identity headers.
It never verifies signatures and never logs the full token (only a 6-char prefix + length). It is off by default and must not be enabled in production. It observes requests only and never affects how they are authenticated.
# Run a product's tests
cd src/vks-mcp-server && uv run pytest tests/ -vConventions for adding tools and new servers live in CLAUDE.md.
Apache-2.0. See LICENSE.
This server cannot be deployed
Deploy, monitor, and manage your OpenClaw AI assistants via natural language.
Provides capabilities that let LLM agents perform a range of infrastructure management tasks.
Manage Rackspace Spot Kubernetes Cloudspaces, node pools, and VMs from your AI assistant.
Your AI Agent's Infrastructure Layer. Connect Claude, Copilot, Codex, or ChatGPT to 200+ managed open source services. Start databases, pipelines, and applications through natural language.
