Provides tools for interacting with Kubernetes clusters to perform Chaos Mesh experiments, including cluster inspection (namespaces, pods, nodes), network chaos, HTTP chaos, pod kill, and chaos lifecycle management (list, delete).
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., "@Chaos Mesh MCP Server Add 200ms network delay between service-a and service-b"
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.
Chaos Mesh MCP 服务端，基于 Python FastMCP 实现，提供 Chaos Mesh 混沌实验的 MCP 工具集。
chaosmesh-mcp/
├── src/
│ └── chaosmesh_mcp/ # 源代码 (src 布局)
│ ├── __init__.py
│ ├── __main__.py # python -m 入口
│ ├── server.py # FastMCP 服务启动
│ ├── tools/ # MCP 工具注册
│ │ ├── chaos.py # list_chaos / delete_chaos
│ │ ├── cluster.py # 集群巡检 (namespace/pod/node)
│ │ ├── httpchaos.py # HTTP 接口级延迟注入
│ │ ├── network_chaos.py # 网络延迟/丢包/外部延迟
│ │ └── pod.py # Pod Kill 注入
│ ├── services/ # 服务适配层
│ │ ├── kubectl.py # kubectl 命令封装
│ │ └── chaosmesh.py # 模板渲染与 Chaos CRD 应用
│ ├── utils/ # 工具函数
│ │ ├── generate.py # 混沌资源名生成
│ │ ├── response.py # 统一响应格式
│ │ └── validate.py # 输入校验
│ └── templates/ # Chaos Mesh YAML 模板
│ ├── network_delay.yaml
│ ├── network_loss.yaml
│ ├── network_external_delay.yaml
│ ├── pod_kill.yaml
│ └── httpchaos_delay.yaml
├── tests/
│ └── __init__.py
├── deploy/k8s/ # Kubernetes 部署清单
│ ├── deployment.yaml
│ ├── service.yaml
│ ├── ingress.yaml
│ ├── kustomization.yaml
│ ├── kustomizeconfig.yaml
│ ├── .env.config
│ └── .env.secret.example
├── pyproject.toml
├── uv.lock
├── .python-version
├── Dockerfile
└── README.mdRelated MCP server: Resilience Architect MCP
# 确保已安装 uv
uv --version
# 同步依赖（自动创建 .venv 并安装所有依赖）
uv sync# 使用 uv run 启动服务
uv run python -m chaosmesh_mcp
# 或通过 console_scripts
uv run chaosmesh-mcpdocker build -t chaos-mcp:latest .
docker run -p 8000:8000 chaos-mcp:latestMCP_HOST — 服务监听地址，默认 0.0.0.0
MCP_PORT — 服务监听端口，默认 8000
tools/cluster.py)get_namespaces — 查询集群中的 namespace 列表
get_pods — 以 wide 格式列出目标命名空间中的 Pod
get_nodes — 以 wide 格式列出集群中的所有节点
cluster_info — 获取 Kubernetes 集群端点与服务信息
tools/network_chaos.py)resolve_namespace — 根据 K8s 服务名称查询所在的 namespace
network_delay — 对服务与服务之间的网络连接注入延迟
network_external_delay — 对发往第三方渠道域名注入响应延迟 (Workflow)
network_loss — 对服务间注入 NetworkChaos 丢包故障
tools/httpchaos.py)httpchaos_delay — 按 HTTP 路径和端口精准注入延迟
tools/pod.py)pod_kill — PodChaos pod-kill 故障注入
tools/chaos.py)list_chaos — 列出指定命名空间中的混沌资源
delete_chaos — 按类型、名称和命名空间删除单个混沌资源
所有工具返回统一的 JSON 响应：
{
"success": true,
"message": "操作描述",
"data": {
"kubectl": "kubectl 返回内容",
"yaml": "渲染后的 YAML"
},
"yaml": "渲染后的 YAML (顶层)"
}kubectl create namespace chaos-mcp
cp deploy/k8s/.env.secret.example deploy/k8s/.env.secret
# 编辑 deploy/k8s/.env.secret 填入真实值
kubectl apply -k deploy/k8s获取 Kubernetes 集群端点与服务信息。
Inputs: no named parameters.
按类型、名称和命名空间删除单个混沌资源。
Inputs: kind (required), name (required), namespace (required).
查询集群中的 namespace 列表。
Inputs: no named parameters.
以 wide 格式列出集群中的所有节点。
Inputs: no named parameters.
以 wide 格式列出目标命名空间中的 Pod。
Inputs: namespace (required).
按 HTTP 路径和端口注入延迟（精准影响指定接口）；如果只需要服务与服务之间的整体网络延迟，请使用 network_delay。
Inputs: namespace (required), app (required), path (required), port, +3 more.
列出指定命名空间中的 PodChaos、NetworkChaos 和 StressChaos 资源。
Inputs: namespace (required).
对服务与服务之间的网络连接注入延迟（支持 to）；不按 HTTP path 匹配。若需按 path 精准注入，请使用 httpchaos_delay。
Inputs: selector_value (required), target_selector_value (required), selector_namespace (required), target_namespace (required), +4 more.
创建 ChaosMesh Workflow，对 发往第三方渠道域名的出方向流量注入响应延迟混沌实验。
Inputs: selector_value (required), selector_namespace (required), external_url (required), duration, +2 more.
对标签选择器匹配到的 Pod 注入 NetworkChaos 丢包故障。
Inputs: selector_value (required), selector_namespace (required), target_selector_value (required), target_namespace (required), +3 more.
对标签选择器匹配到的 Pod 注入 PodChaos pod-kill 故障。
Inputs: namespace (required), selector_value (required), duration.
根据K8s服务名称，查询该服务所在的命名空间namespace
Inputs: selector_value (required).
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 12 tools
Each tool has a clearly distinct purpose. The two delay tools are explicitly disambiguated in their descriptions, and other tools cover different fault types or informational queries without overlap.
Most tools follow a verb_noun pattern (e.g., delete_chaos, list_chaos, pod_kill), but there are minor inconsistencies like 'httpchaos_delay' (combined prefix) and 'network_external_delay' (three-part name) which mildly deviate from the norm.
With 12 tools, the server covers core chaos engineering operations (query, inject, delete) for multiple fault types, without being overwhelming or too sparse.
The set includes common fault types (pod kill, network delay/loss, HTTP delay) and external delay, but lacks stress chaos injection tools despite listing them, and does not support update operations, leaving notable gaps.
Deploy applications and manage governed Kubernetes preview spaces and capacity.
Deploy and operate Kubernetes: clusters, stacks, apps, logs, metrics and security, with OAuth.
Fail-closed policy guardrails for AI agents running kubectl, terraform, helm, and argocd.
Deploy containers on Kubernetes with x402 billing. 9 workload types and source builds.
