Metrxbots

io.github.metrxbots由 GitHub 账号 metrxbots 发布v0.2.2更新于 Oct 24, 2026

Track AI agent costs, detect waste, optimize models, and prove ROI. 23 MCP tools across 10 domains.

概览

AI 生成的概览

Metrx 是一个本地 MCP 服务器,让助手读取 AI 智能体成本数据、发现浪费、比较模型、管理预算并生成 ROI 报告。

可以做什么
读取成本汇总、智能体列表及单个智能体的成本与性能详情,以及活动告警和故障预测,并可确认告警。获取优化建议并应用,按任务推荐模型路由,比较各模型定价。查看、创建和修改预算,包括执行模式以及暂停或恢复。运行模型 A/B 实验、查看结果并停止实验。执行成本泄漏扫描,将任务归因到业务成果,并获取 ROI 与归因报告。设置告警阈值,生成 ROI 审计和升级论证报告。
适用场景
适用于已通过 Metrx 运行 AI 智能体的场景:希望助手在不打开仪表盘的情况下查看支出、排查过度付费或成本失控、测试更便宜的模型,或整理 ROI 与审计报告。
运行要求
通过 npx 或 npm 在本地运行,需要 Node.js。需要 Metrx 账号和 API 密钥,通过 METRX_API_KEY 环境变量提供,或用交互式登录参数保存到本地配置文件;演示模式使用示例数据,无需密钥。需要访问 Metrx API 的网络连接。
安装前请注意
API 密钥是有效凭据,会存放在环境变量或本地配置文件中,需妥善保护。多个工具会更改状态:应用优化、设置或暂停预算、启动和停止实验、确认告警,以及配置带邮件、webhook 或自动暂停的告警阈值。成本与用量数据会发送到 Metrx 服务。

安装

在 SourceWeft 中

  1. 打开 控制台中的 Metrxbots,将其添加到工作区。
  2. 为需要使用其工具的对话启用该服务。

Desktop only,通过 STDIO。 STDIO 服务会启动本地进程,因此需要 SourceWeft 桌面宿主。

其他 MCP 客户端

参照 仓库 中的启动说明。

README

Metrx MCP Server

[npm version] [CI] [License: MIT] [Smithery] [Glama]

Your AI agents are wasting money. Metrx finds out how much, and fixes it.

The official MCP server for Metrx — the AI Agent Cost Intelligence Platform. Give any MCP-compatible agent (Claude, GPT, Gemini, Cursor, Windsurf) the ability to track its own costs, detect waste, optimize model selection, and prove ROI.

[Metrx Server MCP server]

Why Metrx?

ProblemWhat Metrx Does
No visibility into agent spendReal-time cost dashboards per agent, model, and provider
Overpaying for LLM callsProvider arbitrage finds cheaper models for the same task
Runaway costsBudget enforcement with auto-pause when limits are hit
Wasted tokensCost leak scanner detects retry storms, context bloat, model mismatch
Can't prove AI ROIRevenue attribution links agent actions to business outcomes

Quick Start

Try it now — no signup required

bash
npx @metrxbot/mcp-server --demo

This starts the server with sample data so you can explore all 23 tools instantly.

Connect your real data

Option A — Interactive login (recommended):

bash
npx @metrxbot/mcp-server --auth

Opens your browser to get an API key, validates it, and saves it to ~/.metrxrc so you never need to set env vars.

Option B — Environment variable:

bash
METRX_API_KEY=sk_live_your_key_here npx @metrxbot/mcp-server --test

Get your free API key at app.metrxbot.com/sign-up.

Add to your MCP client (Claude Desktop, Cursor, Windsurf)

If you used --auth, no env block is needed — the key is read from ~/.metrxrc automatically:

json
{  "mcpServers": {    "metrx": {      "command": "npx",      "args": ["@metrxbot/mcp-server"]    }  }}

Or pass the key explicitly via environment:

json
{  "mcpServers": {    "metrx": {      "command": "npx",      "args": ["@metrxbot/mcp-server"],      "env": {        "METRX_API_KEY": "sk_live_your_key_here"      }    }  }}

Remote HTTP endpoint

For remote agents (no local install needed):

POST https://metrxbot.com/api/mcpAuthorization: Bearer sk_live_your_key_hereContent-Type: application/json

From npm

bash
npm install @metrxbot/mcp-server

23 Tools Across 10 Domains

Dashboard (3 tools)

ToolDescription
metrx_get_cost_summaryComprehensive cost summary — total spend, call counts, error rates, and optimization opportunities
metrx_list_agentsList all agents with status, category, cost metrics, and health indicators
metrx_get_agent_detailDetailed agent info including model, framework, cost breakdown, and performance history

Optimization (4 tools)

ToolDescription
metrx_get_optimization_recommendationsAI-powered cost optimization recommendations per agent or fleet-wide
metrx_apply_optimizationOne-click apply an optimization recommendation to an agent
metrx_route_modelModel routing recommendation for a specific task based on complexity
metrx_compare_modelsCompare LLM model pricing and capabilities across providers

Budgets (3 tools)

ToolDescription
metrx_get_budget_statusCurrent status of all budget configurations with spend vs. limits
metrx_set_budgetCreate or update a budget with hard, soft, or monitor enforcement
metrx_update_budget_modeChange enforcement mode of an existing budget or pause/resume it

Alerts (3 tools)

ToolDescription
metrx_get_alertsActive alerts and notifications for your agent fleet
metrx_acknowledge_alertMark one or more alerts as read/acknowledged
metrx_get_failure_predictionsPredictive failure analysis — identify agents likely to fail before it happens

Experiments (3 tools)

ToolDescription
metrx_create_model_experimentStart an A/B test comparing two LLM models with traffic splitting
metrx_get_experiment_resultsStatistical significance, cost delta, and recommended action
metrx_stop_experimentStop a running model routing experiment and lock in the winner

Cost Leak Detector (1 tool)

ToolDescription
metrx_run_cost_leak_scanComprehensive 7-check cost leak audit across your entire agent fleet

Attribution (3 tools)

ToolDescription
metrx_attribute_taskLink agent actions to business outcomes for ROI tracking
metrx_get_task_roiCalculate return on investment for an agent — costs vs. attributed outcomes
metrx_get_attribution_reportMulti-source attribution report with confidence scores and top contributors

Alert Configuration (1 tool)

ToolDescription
metrx_configure_alert_thresholdSet cost or operational alert thresholds with email, webhook, or auto-pause

ROI Audit (1 tool)

ToolDescription
metrx_generate_roi_auditBoard-ready ROI audit report for your AI agent fleet

Upgrade Justification (1 tool)

ToolDescription
metrx_get_upgrade_justificationROI report for tier upgrades based on current usage patterns

Prompts

Pre-built prompt templates for common workflows:

PromptDescription
analyze-costsComprehensive cost overview — spend breakdown, top agents, optimization opportunities
find-savingsDiscover optimization opportunities — model downgrades, caching, routing
cost-leak-scanScan for waste patterns — retry storms, oversized contexts, model mismatch

Examples

"How much am I spending?"

User: What was my AI cost this week?→ metrx_get_cost_summary(period_days=7)
Total Spend: $234.56 | Calls: 2,450 | Error Rate: 0.2%├── customer-support: $156.23 (1,800 calls)└── code-generator: $78.33 (650 calls)
💡 Switch customer-support from GPT-4 to Claude Sonnet: Save $42/week

"Find me savings"

User: Am I overpaying for my agents?→ metrx_compare_models(models=["gpt-4o", "claude-3-5-sonnet", "gemini-1.5-pro"])
Model Comparison (per 1M tokens):├── gpt-4o: $2.50 in / $10.00 out├── claude-3-5-sonnet: $3.00 in / $15.00 out└── gemini-1.5-pro: $3.50 in / $10.50 out

"Test a cheaper model"

User: Test Claude 3.5 Sonnet against my GPT-4 setup→ metrx_create_model_experiment(agent_id="agent_123",    model_a="gpt-4o", model_b="claude-3-5-sonnet-20241022", traffic_split=10)
Experiment started: 90% GPT-4o, 10% Claude 3.5 SonnetCheck back in 14 days for statistical significance.

Companion Tool: Cost Leak Detector

This repo also includes @metrxbot/cost-leak-detector — a free, offline CLI that scans your LLM API logs for wasted spend. No signup, no cloud, no data leaves your machine.

bash
npx @metrxbot/cost-leak-detector demo

It runs 7 checks (idle agents, premium model overuse, missing caching, high error rates, context overflow, no budgets, arbitrage opportunities) and gives you a scored report in seconds. See the full docs.

Configuration

API Key (required)

The server looks for your API key in this order:

  1. METRX_API_KEY environment variable
  2. ~/.metrxrc file (created by --auth)

Run npx @metrxbot/mcp-server --auth to save your key, or set the env var directly.

VariableRequiredDescription
METRX_API_KEYYes*Your Metrx API key (get one free)
METRX_API_URLNoOverride API base URL (default: https://metrxbot.com/api/v1)

*Not required if you've run --auth — the key is read from ~/.metrxrc automatically.

CLI Flags

FlagDescription
--demoStart with sample data — no API key or signup needed
--authInteractive login — opens browser, validates key, saves to ~/.metrxrc
--testVerify your API key and connection

Rate Limiting

60 requests per minute per tool. For higher limits, contact [email protected].

Development

bash
git clone https://github.com/metrxbots/mcp-server.gitcd mcp-servernpm installnpm run typechecknpm test

Contributing

See CONTRIBUTING.md for guidelines.

Links

A Note on Naming

The product is Metrx (metrxbot.com). The npm scope is @metrxbot and the Smithery listing is metrxbot/mcp-server. The GitHub organization is metrxbots (with an s) because metrxbot was already taken on GitHub. If you see metrxbot vs metrxbots across platforms, they're the same project — just a GitHub namespace constraint.

License

MIT — see LICENSE.

💬 Feedback

Did Metrx work for you? We'd love to hear it — good or bad.

If you installed but hit a snag, tell us what happened — we read every report.

来源:README.md,提交 fcf7192

工具

0
SourceWeft 尚未在沙箱中启动过该软件包,因此其工具未知。

版本历史

1
  1. v0.2.2最新Sep 16, 2026

评分

已满足适用检查项的 50%

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25 分

    请保持端点能够响应 MCP 握手;SourceWeft 每 30 天探测一次。

  • 已声明认证方式10 分

    请在注册中心条目中声明凭据请求头,或保持端点能够响应探测,以便判定其认证方式。

  • 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