All loops
OpsMedium 8 min· claude-haiku-4-5
Latency budget
Tracks p95 budget per service and pings owners on burn.
Read the full Ops build guide →NOT DEPLOYEDNOT DEPLOYED
0183ms
Trigger
cron(0 7 * * *) fired · every day · 07:00
02783ms
Agent
claude-haiku-4-5 · in 1083 tok · out 377 tok
03313ms
Tools
fetch-mcp/anthropic:messages.create → 429 backoff · 403ms
0473ms
Verify
schema check · json-schema draft-2020 passed
05163ms
Output
pagerduty ack · incident summary posted
0643ms
Notify
audit log written · runbook link attached
SUCCESS
0%
0 runs
P50
0ms
median
P95
0ms
tail
AVG COST
—
per run
LAST OK
never
—
LAST FAIL
never
none
Latency · last 30 runs0 samples
no runs yet
Latest output · what your users see
Datadogcritical
p95 latency on /api/checkout crossed 1200ms for 6 min. Correlated deploy: web@d41f2a. Suggested rollback command attached.
deduped ×9·runbook: runbooks/checkout-latency.md#rollback
// press Test to run once · Watch live to keep streaming · Deploy to make it real
The problem
Ops work is repetitive, urgent, and easy to drop. Humans pay the tax.
The outcome
A scheduled agent that does the chore, posts the result, and escalates only when needed.
Ingredients & skills
Secrets
- ANTHROPIC_API_KEY
Providers
- Anthropic
MCP servers
- fetch-mcp
#ops#sre#sre
How it works
Tracks p95 budget per service and pings owners on burn.
Step 1
1 — Fetch
Read the relevant telemetry. Scope keys to read-only.
Step 2
2 — Analyse
Claude summarises + ranks. Empty input ⇒ no message (no noise).
Step 3
3 — Post
Slack, email, or PagerDuty depending on severity.
One-line deploy
The button above runs the same command with your saved config. This is the raw CLI form.
bash
npx claudeloops deploy latency-budget