AnomalyArmor

ai.anomalyarmor由 anomalyarmor.ai 发布官方发布者v0.6.1更新于 Oct 24, 2026

Data observability tools for engineering teams: alerts, freshness, schema drift, lineage, quality.

已验证Streamable HTTP可网页运行数据库安全与监控数据与分析

概览

AI 生成的概览

让助手查询和管理 AnomalyArmor 数据仓库可观测性：告警、新鲜度、模式漂移、质量规则与血缘。

可以做什么
通过 50 多个工具把 MCP 客户端连接到 AnomalyArmor 账户，覆盖仓库健康与简报、告警与事件、数据源、新鲜度计划、模式监控、数据质量指标与有效性规则、告警目标、血缘、标签以及异步任务。只读工具用于列出和汇总告警、新鲜度、漂移和质量情况；写入工具用于创建或更新规则、计划、指标、数据源和目标。还提供自然语言问答与 AI 分析，耗时任务以可轮询的任务 ID 返回。
适用场景
当团队已使用 AnomalyArmor 监控数据仓库，并希望助手以对话方式处理告警、检查新鲜度或模式漂移、查看血缘或调整监控规则，而不是打开网页界面时，适合使用。
运行要求
可使用托管的远程端点（Streamable HTTP，OAuth 2.1），也可在本地运行 armor-mcp 包（PyPI，stdio），通过 uvx 或 pip 启动。本地方式需要 AnomalyArmor API 密钥，以 ARMOR_API_KEY 环境变量或本地配置文件提供，并需要访问 AnomalyArmor 服务的网络。
安装前请注意
本地方式需要密钥 ARMOR_API_KEY，其权限范围限于你所在公司的数据。许多工具会写入：create、update、manage、enable、disable 会改动告警规则、计划、指标、数据源和目标，删除或忽略类操作具有破坏性。generate_intelligence 与 trigger_asset_discovery 会排队执行付费任务，运行前请确认。告警数据可能被发送到 Slack、邮件、Webhook 或 PagerDuty 等第三方目标。

安装

在 SourceWeft 中

  1. 打开 控制台中的 AnomalyArmor，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.anomalyarmor%2Farmor-mcp/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": {
    "anomalyarmor": {
      "type": "http",
      "url": "https://mcp.anomalyarmor.ai/mcp"
    }
  }
}

README

armor-mcp

[PyPI] [License: MIT] [MCP] [Glama]

The AnomalyArmor MCP server lets AI assistants (Claude Code, Cursor, Claude Desktop, any MCP client) interact with your data warehouse through 50+ structured tools: query alerts, monitor freshness, investigate schema drift, configure validity rules, and trace lineage in natural language.

Two ways to connect

MethodAuthBest for
Remote (recommended)OAuth 2.1 (Clerk)Zero install, always up to date
LocalAPI keyAir-gapped, on-call laptops, custom proxies

Remote (recommended)

The server is hosted at https://mcp.anomalyarmor.ai/mcp (Streamable HTTP transport). First call opens a browser for OAuth; the token is cached for ~12h.

Claude Code

bash
claude mcp add anomalyarmor --transport http https://mcp.anomalyarmor.ai/mcp

Cursor

Add to ~/Library/Application Support/Cursor/mcp.json (macOS), ~/.config/Cursor/mcp.json (Linux), or %APPDATA%\Cursor\mcp.json (Windows):

json
{  "mcpServers": {    "anomalyarmor": {      "url": "https://mcp.anomalyarmor.ai/mcp"    }  }}

Any MCP client

URL: https://mcp.anomalyarmor.ai/mcpTransport: streamable-httpAuth: OAuth 2.1Discovery: https://mcp.anomalyarmor.ai/.well-known/oauth-protected-resource

Local (API key)

Use when the remote server isn't reachable or when you want no third-party dependency.

Install

bash
# Recommended: uvx (no install)uvx armor-mcp
# Or pippip install armor-mcp

Configure

Get an API key from Settings → API Keys at https://app.anomalyarmor.ai.

bash
export ARMOR_API_KEY=aa_live_your_key_here

Or persist to ~/.armor/config.yaml:

yaml
api_key: aa_live_your_key_here

Wire to a client

json
{  "mcpServers": {    "anomalyarmor": {      "command": "uvx",      "args": ["armor-mcp"],      "env": {        "ARMOR_API_KEY": "aa_live_your_key_here"      }    }  }}

Tools

53 tools across 8 categories. Every tool ships with MCP ToolAnnotations (readOnlyHint, destructiveHint, idempotentHint, openWorldHint: false) so hosts can decide when to auto-confirm.

Health & Briefings

ToolDescription
health_summaryOverall warehouse health: stale assets, drift, alerts
get_todays_briefingDaily digest with key insights
get_coverageMonitoring coverage analysis
manage_coverageUpdate coverage targets
recommendAI-suggested next monitoring actions

Alerts & Incidents

ToolDescription
list_alertsQuery alerts with filters (severity, date, status)
list_inbox_alertsAlerts not yet triaged
get_alerts_summaryAggregate counts and trends
get_alert_trendsPeriod-over-period trend analysis
get_alert_historyActivity timeline for an alert
update_alertAcknowledge / resolve / dismiss / snooze
list_alert_rulesView configured rules
create_alert_ruleAdd a new alert rule
manage_alert_ruleUpdate / delete / preview rules
manage_rule_destinationsWire rules to destinations

Assets

ToolDescription
list_assetsList data sources with filters
create_assetConnect a new source
manage_assetUpdate / delete / test connection
trigger_asset_discoveryStart schema discovery (async)

Freshness

ToolDescription
get_freshness_summaryFreshness overview
check_freshnessCheck a specific asset / table
setup_freshnessBulk-create freshness schedules
list_freshness_schedulesView configured schedules
manage_freshness_scheduleUpdate / delete a schedule

Schema Monitoring

ToolDescription
get_schema_summaryDrift overview
list_schema_changesRecent changes with severity
create_schema_baselineCapture current schema as a baseline
enable_schema_monitoringStart drift detection
disable_schema_monitoringStop drift detection

Data Quality (Metrics & Validity)

ToolDescription
get_metrics_summaryMetric health by asset
list_metricsList configured metrics
create_metricAdd a metric (row count, null %, etc.)
manage_metricUpdate / delete / capture a metric
get_validity_summaryPass/fail rate for validity rules
list_validity_rulesList validity rules
create_validity_ruleAdd a NOT_NULL / UNIQUE / RANGE / REGEX rule
manage_validity_ruleUpdate / delete / run a rule
create_referential_checkCross-table referential check
manage_referentialUpdate / delete a referential check

Destinations & Routing

ToolDescription
list_destinationsSlack / email / webhook / PagerDuty configs
setup_destinationAdd a new destination
manage_destinationUpdate / delete / test

Intelligence, Lineage, Tags, Jobs

ToolDescription
ask_questionSynchronous natural-language Q&A about your data
generate_intelligenceTrigger AI analysis (async, expensive)
get_lineageUpstream/downstream dependency graph
list_tagsTags for an asset
create_tagTag a table or column
apply_tagsBulk tag application
job_statusTrack an async job
cancel_jobCancel a running async job
get_api_key_infoInspect the current API key's scope

Full descriptions, parameter schemas, and example prompts: https://docs.anomalyarmor.ai/integrations/mcp-server.

Tool safety conventions

Hosts respect the annotations:

CategoryPatternConfirmation
Read-onlylist_*, get_*, check_*, search_*Auto-invoke
Mutatingcreate_*, update_*, enable_*, disable_*, manage_*Confirm with user
Destructivedelete_* (via manage_*), dismiss_alertAlways confirm; list affected entities
Expensivegenerate_intelligence, trigger_asset_discoveryConfirm; these queue paid jobs

Multi-tenancy

The OAuth token (or API key) is scoped to the user's company. You cannot impersonate another tenant. Don't fabricate company_id arguments.

Async jobs

trigger_asset_discovery, generate_intelligence, and metric captures return a job_id. Poll with job_status(job_id) rather than re-triggering.

Discovery surfaces

FilePurpose
https://mcp.anomalyarmor.ai/.well-known/oauth-protected-resourceRFC 9728 protected-resource metadata
https://www.anomalyarmor.ai/.well-known/mcp/server-card.jsonSEP-1649 MCP server card
https://www.anomalyarmor.ai/.well-known/agent-card.jsonA2A v0.2 agent card
https://www.anomalyarmor.ai/.well-known/api-catalogRFC 9727 API catalog linkset

Development

bash
git clone https://github.com/anomalyarmor/agents.gitcd agents/armor-mcpuv syncuv run pytest

License

MIT — see LICENSE.

来源：armor-mcp/README.md，提交 74fed02

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. v0.6.1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