Rius

ai.glassflow由 glassflow.ai 发布v0.12.0更新于 Oct 27, 2026

Query your AI agents' traces, errors, cost and alerts in Rius, the agent observability platform.

已验证Streamable HTTP可网页运行安全与监控数据与分析AI 与机器学习

概览

AI 生成的概览

托管式 Rius MCP 服务器，让助手查询 Rius 可观测性平台中 AI 智能体的追踪、错误、成本、指标与告警。

可以做什么
  • 查看工作区与身份：当前用户、工作区、用户及用户活动摘要
  • 浏览智能体：列出智能体、版本、版本对比与生命周期事件
  • 探索追踪与跨度：追踪摘要、追踪列表、单条追踪、跨度与搜索目录
  • 查看错误与指标：错误分组、工作区指标概览与指标目录
  • 读取与管理告警：列出告警、规则与渠道，并创建、更新、删除规则和渠道以及确认告警
  • 为工作区创建 API 密钥
适用场景
当你的 AI 智能体已向 Rius 上报追踪数据，并希望在编码助手中直接询问失败、成本、延迟、版本差异或未处理告警时适用。也适合让助手配置或调整告警规则的团队。
运行要求
远程 Streamable HTTP 端点，无需安装或自托管。需要一个已有智能体追踪数据的 Rius 工作区，以及 OAuth 2.1 浏览器登录，或在无浏览器场景下通过 Authorization 头以 bearer 令牌方式提供 Rius API 密钥（例如使用 RIUS_API_KEY）。仅支持 stdio 的客户端需借助 mcp-remote 之类的桥接。
安装前请注意
写入类工具会改动数据：可创建、更新或删除告警规则与渠道，确认告警，并创建 API 密钥。登录状态下告警与密钥工具需要组织管理员权限；API 密钥默认只读，除非授予 alerts:manage 或 channels:manage，且确认告警与创建密钥在密钥模式下始终不可用。请将 API 密钥视为机密，避免粘贴到共享配置中。

安装

在 SourceWeft 中

  1. 打开 控制台中的 Rius，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/ai.glassflow%2Frius/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": {
    "rius": {
      "type": "http",
      "url": "https://mcp.eu.console.rius-glassflow.com/mcp"
    }
  }
}

README

Rius MCP server

[Rius MCP connector]

Rius is an observability platform for AI agents, built by GlassFlow. The Rius MCP server lets your coding assistant query your agents' traces, errors, cost and alerts in plain language, from VS Code (GitHub Copilot), Claude Code, Cursor, Codex or any other MCP client.

This repository holds the public README for the hosted server, plus a reference copy of its server.json. The live registry entry is published by the server's release pipeline, so its version may be newer than the copy here. The server itself runs as a managed service; there is nothing to install or self-host.

  • Endpoint: https://mcp.eu.console.rius-glassflow.com/mcp
  • Transport: Streamable HTTP
  • Auth: OAuth 2.1 sign-in in the browser, or a Rius API key for headless use
  • MCP registry name: ai.glassflow/rius
  • Docs: https://docs.glassflow.ai/rius/mcp

What you can ask

Once connected, ask questions about your agents in the chat:

  • "Which of my agents had the most errors in the last 24 hours, and what are the top error groups?"
  • "Show me today's failed runs for support-agent and walk me through the worst one."
  • "Compare cost and latency between versions 1.4.0 and 1.5.0 of support-agent."
  • "Which end users cost the most this week?"
  • "List open alerts and explain what triggered each one."
  • "Create an alert that fires when checkout-agent error rate goes above 5%."

Prerequisites

A Rius workspace with agent traces in it. Sign up and follow the quickstart to send your first traces with the Rius SDK or any OpenTelemetry exporter.

Connect

Every client below uses the server name rius. On first use the client signs you in to Rius through the browser (some clients need a command to start it; see the per-client notes). The server supports dynamic client registration and client ID metadata documents, so there is no client ID to configure.

VS Code (GitHub Copilot)

Add the server to .vscode/mcp.json in your workspace (or your user mcp.json). Note the top-level key is servers:

json
{  "servers": {    "rius": {      "type": "http",      "url": "https://mcp.eu.console.rius-glassflow.com/mcp"    }  }}

Start the server from the Start action above its entry in mcp.json (or run MCP: List Servers) and allow the sign-in prompt. Then open Copilot Chat, switch to Agent mode and select the rius tools.

Claude Code

bash
claude mcp add --transport http rius https://mcp.eu.console.rius-glassflow.com/mcp

Run /mcp inside Claude Code to sign in.

Cursor

Add to ~/.cursor/mcp.json (or .cursor/mcp.json in a project):

json
{  "mcpServers": {    "rius": {      "url": "https://mcp.eu.console.rius-glassflow.com/mcp"    }  }}

Then approve the server under Settings → MCP.

Codex

bash
codex mcp add rius --url https://mcp.eu.console.rius-glassflow.com/mcp

Other clients

Most clients accept a JSON block like this one:

json
{  "mcpServers": {    "rius": {      "type": "http",      "url": "https://mcp.eu.console.rius-glassflow.com/mcp"    }  }}

Clients that only support stdio can bridge through mcp-remote:

bash
npx mcp-remote https://mcp.eu.console.rius-glassflow.com/mcp

Per-client notes (Claude Desktop, Windsurf, Cline, Zed, Gemini CLI, opencode) are in the MCP server docs.

Headless use with an API key

For CI jobs, scripts or remote agents that cannot open a browser, send a Rius API key as a bearer token instead of signing in. Create a key with the Query data capability from the console's MCP page ("Connect an agent") or the Create API key dialog.

bash
export RIUS_API_KEY=...claude mcp add --transport http rius https://mcp.eu.console.rius-glassflow.com/mcp \  --header "Authorization: Bearer ${RIUS_API_KEY}"

In VS Code, let it prompt for the key once and store it securely:

json
{  "inputs": [    {      "type": "promptString",      "id": "rius-key",      "description": "Rius API key",      "password": true    }  ],  "servers": {    "rius": {      "type": "http",      "url": "https://mcp.eu.console.rius-glassflow.com/mcp",      "headers": {        "Authorization": "Bearer ${input:rius-key}"      }    }  }}

In other JSON configs, add a headers block. Not every client expands environment variables, so use your client's variable syntax or paste the key:

json
{  "mcpServers": {    "rius": {      "type": "http",      "url": "https://mcp.eu.console.rius-glassflow.com/mcp",      "headers": {        "Authorization": "Bearer ${RIUS_API_KEY}"      }    }  }}

In Codex, set bearer_token_env_var = "RIUS_API_KEY" on the server entry in ~/.codex/config.toml.

Tools

The main tools are below. The full, current reference with parameters is at https://docs.glassflow.ai/rius/mcp/tools.

AreaTools
Workspace and identityget_me, list_workspaces, list_users, user_activity_summary
Agentslist_agents, agent_versions, compare_agent_versions, agent_lifecycle_events
Traces and spansagent_traces_summary, list_agent_traces, get_agent_trace, list_spans, describe_search_catalog
Errors and metricsagent_error_groups, workspace_metrics_overview, list_metrics_catalog
Alerts (read)list_alerts, get_alert, list_alert_rules, list_alert_channels, describe_alert_catalog
Alerts (write)create_alert_rule, update_alert_rule, delete_alert_rule, create_alert_channel, update_alert_channel, delete_alert_channel, ack_alert
Keys (write)create_api_key

The tools marked "write" change data. When a call is not allowed, the server tells the assistant why instead of failing silently:

  • Signed in: read tools and ack_alert work for every member. Alert rule and channel tools, list_alert_channels and create_api_key need an org admin.
  • With an API key: a key is read-only unless it was created with alerts:manage (alert rule tools) or channels:manage (alert channel tools, including list_alert_channels). ack_alert and create_api_key never work with a key.

Links

License

The contents of this repository are released under the MIT License. The Rius service is governed by its own terms.

来源：README.md，提交 e3ba244

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

2
  1. v0.12.0最新Oct 22, 2026
  2. v0.11.2Oct 21, 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