Causely

app.causely由 causely.app 发布官方发布者v1.0.1更新于 Oct 24, 2026

Causal reasoning for reliability: root cause, blast radius, and service health for agents.

已验证Streamable HTTP可网页运行云端与基础设施安全与监控

概览

AI 生成的概览

托管式 Causely MCP 服务器，让助手在可靠性排查中查询因果模型，获取根因、影响范围和服务健康状况。

可以做什么
  • 查询 Causely 因果模型，对故障进行确定性的根因分析
  • 以结构化结果查看影响范围（blast radius）与服务健康状况
  • 支持告警分诊、变更影响、健康报告、Kubernetes 排查、关联事件与事后复盘等排查流程
  • 借助随附的技能包与常见 MCP 客户端起始配置，在 Causely 工具之间路由任务
适用场景
当助手需要基于 Causely 的因果模型排查生产故障或回答可靠性问题，而不是扫描原始遥测数据时适用；也适合已经在使用 Causely 做可观测性的团队。
运行要求
使用 Causely 托管的远程 MCP 端点，无需本地包或运行时。需要 Causely 账号：首次连接时客户端会打开浏览器进行 OAuth 登录。托管服务器不需要 API 密钥或自定义请求头。
安装前请注意
连接后助手可访问你的 Causely 账号数据，包括服务健康状况与故障上下文，这些数据会发送到 Causely 托管服务。登录通过 OAuth 浏览器流程完成，请确认所授权的账号与组织。该仓库还包含会改动集群状态的 Kubernetes 安装 CLI 与 API 脚本，它们与 MCP 服务器相互独立。

安装

在 SourceWeft 中

  1. 打开 控制台中的 Causely，将其添加到工作区。
  2. 连接 所需的 凭据。它们会被私密存储，绝不会在公开页面上展示。
  3. 为需要使用其工具的对话启用该服务。

Web executable，通过 Streamable HTTP。 远程服务在工作区中配置后即可从网页运行时运行。

交给智能体安装

把这段提示词粘贴到 Claude Code、Codex、Cursor 等智能体，或 SourceWeft 对话中。智能体会读取该服务器的安装指南，先向你展示发布者、端点和登录方式，经你同意后再把它添加到你的客户端。密钥不会写进任何命令。

阅读 https://sourceweft.com/mcp/app.causely%2Fcausely/install.md，并按其中说明把这个 MCP 服务器接入我的客户端。修改任何配置之前，先告诉我它的发布者、端点和登录方式，等我确认。不要把 API 密钥写进命令或我会提交的文件；问我密钥应从哪里获取。

查看智能体所依据的安装指南

其他 MCP 客户端

把它添加到你客户端的 mcpServers 配置中。 客户端会在首次连接时完成 OAuth 登录。

{
  "mcpServers": {
    "causely": {
      "type": "http",
      "url": "https://api.causely.app/mcp"
    }
  }
}

README

Causely Agent Integrations — MCP Skills, Plugins & CLI

These are the tools that let AI agents query Causely's causal model directly, instead of scanning raw telemetry and guessing. This repository bundles client-side resources for working with Causely: agent integrations (MCP), a Kubernetes install CLI, and a shell toolkit for the Causely GraphQL API.

The Causely MCP server is hosted at https://api.causely.app/mcp. Point any MCP-compatible client at it and sign in with your Causely account on first connect (OAuth, no API keys). It exposes Causely's causal model, deterministic root cause, blast radius, and service health, as structured tools your agent can call directly.

AreaPathWhat it is
MCP skills & pluginsmcp/Claude (and compatible) skills that route work across Causely's MCP tools, plus starter configs for Cursor, Claude, Codex, VS Code Copilot, and OpenCode.
Kubernetes CLIcli/Go CLI that wraps Helm to install and manage the Causely agent in-cluster.
API shell clientapi/Bash libraries and scripts for snapshots, comparisons, and CI workflows against the Causely API.

Product documentation lives at docs.causely.ai. For MCP authentication, headers, and troubleshooting, start with MCP Server Integration.

MCP (mcp/)

Point any MCP-compatible client directly at the hosted server:

https://api.causely.app/mcp

On first connect, your client opens a browser to sign in to Causely (OAuth). No API keys or custom headers are required for the hosted server.

Or install the pre-built skills below for Claude Code, Claude Desktop, and Cursor, so your agent gets the right investigation sequence without hand-writing tool-selection logic.

  • mcp/README.md — Overview of the seven packaged skills (alert triage, change impact, health reporting, K8s investigation, correlated incidents, postmortems, and the master MCP router).
  • mcp/plugins/README.md — How to wire the hosted MCP server into Claude, Cursor, Codex, GitHub Copilot, OpenCode, and related tools, including optional custom headers.
  • Leaf skills — six drop-in folders under mcp/skills/, each with SKILL.md and references/complete-investigation.md (symlink to the shared mcp/complete-investigation.md in this repo).
  • Master router skill — mcp/SKILL.md (same causely-mcp role as in the skills overview) plus mcp/complete-investigation.md at the mcp/ root for full tool routing and investigation patterns.

Curious why causal context matters for agent accuracy? See the benchmark results — 100% fault scenario accuracy and 48% fewer tokens per investigation across 72 experiments.

CLI (cli/)

Install and operate the Causely agent from the terminal. See cli/README.md for the install script, causely agent install, auth, and Helm-oriented flags. Official install guide: CLI Installation.

bash
go build -C cli -o causely ../causely version

API client (api/)

Shell-based GraphQL client: snapshots, diffing, GitHub Actions examples, and numbered docs. See api/README.md and api/docs/.

Contributing

See CONTRIBUTING.md for project layout, development setup, and how to submit changes.

The Go module for the CLI is declared in go.mod at the repository root (github.com/Causely/causely-api-client).

来源：README.md，提交 f8b9664

工具

0
该服务器需要登录，因此 SourceWeft 无法在没有账号的情况下读取其工具，也未收录。已连接账号所能看到的内容可能因账号及其权限而异。

版本历史

1
  1. v1.0.1最新Sep 16, 2026

评分

已满足适用检查项的 85%

该评分衡量此收录条目的完整程度，依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价，也不表示服务器安全或当前可用：可连通性是带日期的检查结果，评分也不包含安装量或用户评价。

  • 近 30 天内可连通必需25 分
  • 已声明认证方式必需10 分
  • 已列出工具15 分

    请让服务器在无凭据时响应 tools/list，或发布可在沙箱中无凭据启动的软件包。

  • 描述必需15 分
  • 图标10 分
  • 文档15 分
  • 许可证10 分