Now in public beta

Your AI agents,
fully observable

Monitor every MCP tool call in real-time. Track costs, catch errors, replay sessions. One line of code to instrument your agents.

Start free → See how it works
$ npm install mcp-monitor
$ monitor.wrap(server) // that's it
app.mcpmonitor.io
MCP Monitor Overview Sessions Alerts Raw Logs
All systems normal
LIVE
Total Calls
0
Error Rate
0%
0 failed
Estimated Cost
$0.00
Avg Latency
0ms
ACTIVITY
HEALTH
100%
success
0 OK 0 ERR
AGENTS
agent-compta
0 calls
0ms
agent-ventes
0 calls
0ms
agent-support
0 calls
0ms
LIVE LOGS
ALERTS
All systems normal
Features

Everything you need to
trust your agents

From real-time monitoring to intelligent alerts, MCP Monitor gives you complete visibility into your AI agents' behavior.

Real-time Dashboard

See every tool call as it happens. Live activity charts, agent performance, cost tracking — all in one place.

Session Replay

Replay any agent session step by step. Watch the flow, understand decisions, spot the exact moment things went wrong.

Smart Alerts

Get notified when agents loop, error rates spike, or costs exceed thresholds. Catch issues before your users do.

Cost Tracking

Track estimated costs per tool call, per session, per agent. Know exactly where your money goes.

Deep Inspection

Click any log to see full parameters, responses, errors, and timing. Search across all your data instantly.

One-line Setup

Install the npm package, wrap your MCP server, done. Zero config, zero overhead. Works with any MCP server.

Why it matters

Real problems.
Caught in real-time.

Without monitoring, these issues go unnoticed for hours. With MCP Monitor, you catch them in seconds.

Agent loop detected
Your accounting agent called get_balance 47 times in 30 seconds. MCP Monitor caught the loop and alerted you before it burned through your budget.
Cost spike caught
A single session spent $0.85 instead of the usual $0.02. Session Replay showed the agent was retrying a failed send_email tool with increasingly long prompts.
Silent failure exposed
Your support agent's search_customer tool was returning empty results for 2 hours. No crash, no error — just wrong data. Deep Inspection revealed a malformed query parameter.
How it works

Three lines to full
observability

No complex setup. No infrastructure. Just install, wrap, and monitor.

1

Install the SDK

Add MCP Monitor to your project with npm.

$ npm install mcp-monitor
2

Wrap your server

Import the monitor and wrap your MCP server. All tool calls are automatically intercepted.

import { createMonitor } from 'mcp-monitor'

const monitor = createMonitor({
  apiKey: 'mk_your_key',
  agent: 'my-agent'
})

monitor.wrap(server) // done!
3

Watch it live

Open your dashboard and see every tool call, error, and cost in real-time. Replay sessions, set up alerts, and understand your agents.

Pricing

Start free, scale as you grow

Generous free tier for getting started. Upgrade when you need more.

Free
$0/month
Perfect for testing and small projects
  • 5,000 logs per month
  • 24-hour data retention
  • 2 alert rules
  • 1 agent
  • Session replay
  • Community support
Get started free
Pro
$39/month
For teams running agents in production
  • 50,000 logs per month
  • 30-day data retention
  • Unlimited alert rules
  • Unlimited agents
  • Session replay
  • Priority support
Start free trial →

Your agents made 1,000 tool calls today.
Do you know which ones failed?

Set up monitoring in 30 seconds. Free forever up to 5,000 logs.

Start monitoring free →