AgentWatch runs on your own machine — the client starts it, so there is no endpoint to ping. 62 installs a week from npm. Last commit 16 Apr 2026.
Multi-agent observability: cascade failure detection, heartbeats, and forensic replay
Nothing serious here today
Today is the operative word: we check AgentWatch every 15 minutes and re-read its code on every release. Watch it and you find out the day that stops being true.
Connect this server
This server runs on your own machine — install it with the package manager and the client starts it for you. Package name taken from the official registry entry.
claude mcp add agentwatch -- npx -y @nicofains1/agentwatch{
"mcpServers": {
"agentwatch": {
"command": "npx",
"args": [
"-y",
"@nicofains1/agentwatch"
]
}
}
}[mcp_servers.agentwatch]
command = "npx"
args = ["-y", "@nicofains1/agentwatch"]{
"mcpServers": {
"agentwatch": {
"command": "npx",
"args": [
"-y",
"@nicofains1/agentwatch"
]
}
}
}{
"servers": {
"agentwatch": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@nicofains1/agentwatch"
]
}
}
}Optional settings from the registry entry, left out of the block above:
AGENTWATCH_DB (Path to SQLite database file).
Alternatives to AgentWatch
same job, measured the same wayAI agent observability with deterministic record/replay for debugging agent failures.
Agent tracing, cost tracking, anomaly detection for LLM agents
Unified observability gateway for AI agents — Prometheus, Loki & more, with anomaly detection.
Resumable API pagination, repeated-failure detection and tool-progress cache hints for AI agents
Secure, self-hosted New Relic observability and administration tools for MCP clients
Deterministic fault-injection server for testing how MCP clients handle failures and recover.
Canonical Agentry discovery for coding agents: analytics, errors, deploys, and automation.
Security scanning and threat detection for AI agents
AgentWatch — questions
Answers built from our own checks of this server.