Overview
AgentPulse is a hosted observability platform for AI agents. Report each run with a single HTTP call and get live dashboards, alert rules, cost-by-product analytics, and execution traces.
What is AgentPulse?
AgentPulse answers one question reliably: is anything wrong with my agents right now? Agents fail quietly — a retry loop, a degraded model provider, a prompt that silently doubled in size — and the usual way you find out is the monthly invoice. AgentPulse is built to tell you the moment it happens instead.
You report each agent run with one authenticated HTTP call. AgentPulse stores it, updates your live metrics, attributes its cost to the right product, and evaluates your alert rules — firing to Slack, a webhook, or email when a threshold is breached. There is no SDK to install and no agent code to change.
How it works
Key capabilities
Live dashboard
Run-volume charts, fleet health, success rate, average latency, and active-agent count, with a 24h / 7d / 30d period selector across every view.
Alert rules
Five metrics — error_rate, success_rate, avg_latency_ms, failed_runs, and cost_usd — with gt/lt thresholds and rolling windows. See Alert Rules.
Cost by product
USD spend attributed to each product with its share of the total, so a runaway agent shows up as a bar that does not belong. See Cost by Product.
Execution traces
A step-by-step breakdown of every run — LLM calls, tool calls, retrievals — with per-step timing and tokens. See Execution Traces.
Multi-tenant orgs
Full isolation per organisation, with admin/member roles and products scoped to the org. See Products & Agents.
Hosted & managed. AgentPulse is a hosted SaaS. There is nothing to deploy or operate — sign up at agentpulse.sentinelflux.in and start sending runs.
Where to go next
- Quick Start — register an agent and send your first run in minutes
- Ingestion API — the full
POST /v1/runsreference - Alert Rules — catch failures before they cost you