Implement observability for Lindy AI integrations. Use when setting up monitoring, logging, tracing, or building dashboards for Lindy operations. Trigger with phrases like "lindy monitoring", "lindy observability", "lindy metrics", "lindy logging", "lindy tracing".
npxskills add jeremylongshore/claude-code-plugins-plus-skills--skill lindy-observabilityLoading…
Monitor Lindy AI agent execution health, task completion rates, step-level failures, trigger frequency, and credit consumption. Lindy provides built-in task history in the dashboard. External observability requires webhook callbacks, the Task Completed trigger, and application-side metrics collection.
| Signal | Source | Why It Matters |
|---|---|---|
| Task completion rate | Tasks tab / callback | Measures agent reliability |
| Task duration | Task detail view | Tracks performance over time |
| Step failure rate | Task detail (red steps) | Identifies broken actions |
| Credit consumption | Billing dashboard | Budget tracking |
| Trigger frequency | Task count over time | Detects trigger storms |
| Agent error rate |
| Failed tasks / total tasks |
| Overall health indicator |
Lindy's Tasks tab provides per-agent monitoring:
Use Lindy's built-in Task Completed trigger to build an observability agent:
Monitoring Agent:
Trigger: Task Completed (from Production Support Agent)
Condition: "Go down this path if the task failed"
→ Action: Slack Send Channel Message to #ops-alerts
Message: "Support Agent task failed: {{task.error}}"
Condition: "Go down this path if task duration > 30 seconds"
→ Action: Slack Send Channel Message to #ops-alerts
Message: "Support Agent slow: {{task.duration}}s"
Configure agents to call your metrics endpoint on task completion:
// metrics-collector.ts — Receive agent metrics via HTTP Request action
import express from 'express';
import { Counter, Histogram, Gauge } from 'prom-client';
const app = express();
app.use(express.json());
// Prometheus metrics
const taskCounter = new Counter({
name: 'lindy_tasks_total',
help: 'Total Lindy agent tasks',
labelNames: ['agent', 'status'],
});
const taskDuration = new Histogram({
name: 'lindy_task_duration_seconds',
help: 'Lindy task execution duration',
labelNames: ['agent'],
buckets: [1, 2, 5, 10, 30, 60, 120],
});
const creditGauge = new Gauge({
name: 'lindy_credits_consumed',
help: 'Credits consumed per task',
labelNames: ['agent'],
});
// Receive metrics from Lindy HTTP Request action
app.post('/lindy/metrics', (req, res) => {
const auth = req.headers.authorization;
if (auth !== `Bearer ${process.env.LINDY_WEBHOOK_SECRET}`) {
return res.status(401).json({ error: 'Unauthorized' });
}
const { agent, status, duration, credits } = req.body;
taskCounter.inc({ agent, status });
taskDuration.observe({ agent }, duration);
creditGauge.set({ agent }, credits);
res.json({ recorded: true });
});
// Prometheus scrape endpoint
app.get('/metrics', async (req, res) => {
res.set('Content-Type', 'text/plain');
res.send(await register.metrics());
});
Lindy agent configuration: Add an HTTP Request action as the last step in each monitored agent:
URL: https://monitoring.yourapp.com/lindy/metrics
Method: POST
Body (Set Manually):
{
"agent": "support-bot",
"status": "{{task.status}}",
"duration": "{{task.duration}}",
"credits": "{{task.credits}}"
}
Key panels for a Lindy monitoring dashboard:
| Panel | Metric | Type |
|---|---|---|
| Task Success Rate | rate(lindy_tasks_total{status="completed"}[1h]) | Percentage gauge |
| Task Failures | rate(lindy_tasks_total{status="failed"}[1h]) | Counter |
| Duration p50/p95 | histogram_quantile(0.95, lindy_task_duration_seconds) | Time series |
| Credit Burn Rate | rate(lindy_credits_consumed[1h]) | Counter |
| Active Agents | Count of agents with tasks in last 24h | Stat panel |
| Trigger Frequency | Tasks per hour by agent | Bar chart |
# Prometheus alert rules
groups:
- name: lindy
rules:
- alert: LindyAgentHighFailureRate
expr: rate(lindy_tasks_total{status="failed"}[30m]) > 0.1
for: 10m
labels:
severity: warning
annotations:
summary: "Lindy agent {{ $labels.agent }} failure rate > 10%"
- alert: LindyAgentDown
expr: absent(lindy_tasks_total{agent="support-bot"}[1h])
for: 30m
labels:
severity: critical
annotations:
summary: "No tasks from support-bot in 1 hour"
- alert: LindyCreditsBurnRate
expr: rate(lindy_credits_consumed[1h]) * 720 > 5000
for: 15m
labels:
severity: warning
annotations:
summary: "Credit burn rate will exhaust monthly budget"
Use Lindy Evals to catch quality regressions:
Click the test tube icon below any agent step
Define scoring criteria (LLM-as-judge):
Score 1 (pass) if the response is professional, accurate, and under 200 words.
Score 0 (fail) if the response contains hallucinations or exceeds 200 words.
Run evals against historical task data
Track scores over time to detect quality drift
Note: Eval runs consume credits but do NOT execute real actions (safe simulation).
| Level | What You Monitor | How |
|---|---|---|
| L0 | Nothing | Manual dashboard checks |
| L1 | Task failures | Task Completed trigger + Slack alerts |
| L2 | Success rate + duration | HTTP Request action + Prometheus |
| L3 | Credit burn + quality | Evals + Grafana dashboards |
| L4 | Automated remediation | Monitoring agent auto-restarts failed agents |
| Issue | Cause | Solution |
|---|---|---|
| Metrics endpoint down | Monitoring server crashed | Alert on scrape failures |
| Task Completed not firing | Monitoring agent paused | Check monitoring agent is active |
| Credit burn alert false positive | Legitimate traffic spike | Tune alert threshold |
| Eval scores dropping | Prompt drift or model change | Review recent prompt/model changes |
Proceed to lindy-incident-runbook for incident response procedures.
Create or update AgentSkills. Use when designing, structuring, or packaging skills with scripts, references, and assets.
Create or update AgentSkills. Use when designing, structuring, or packaging skills with scripts, references, and assets.
Set up and use 1Password CLI (op). Use when installing the CLI, enabling desktop app integration, signing in (single or multi-account), or reading/injecting/running secrets via op.
CLI to manage emails via IMAP/SMTP. Use `himalaya` to list, read, write, reply, forward, search, and organize emails from the terminal. Supports multiple accounts and message composition with MML (MIME Meta Language).
Create or update AgentSkills. Use when designing, structuring, or packaging skills with scripts, references, and assets.
Create or update AgentSkills. Use when designing, structuring, or packaging skills with scripts, references, and assets.
Set up and use 1Password CLI (op). Use when installing the CLI, enabling desktop app integration, signing in (single or multi-account), or reading/injecting/running secrets via op.
CLI to manage emails via IMAP/SMTP. Use `himalaya` to list, read, write, reply, forward, search, and organize emails from the terminal. Supports multiple accounts and message composition with MML (MIME Meta Language).