FIG 0.1
Built for ownership
Deploy to your infrastructure and keep control of your operational data
Latest update · v0.3.2 released
Own your uptime
Pulse monitors your services, opens incidents,
and tracks the
providers behind them
Overview
All systems
| Status | Monitor | Uptime 24h | Timeline | Latency | Last Checked |
|---|
Dependencies
Provider reported
Operational
OpenAI APIOpenAI
Operational
Stripe APIStripe
Checking
Vercel RuntimeVercel
FIG 0.2
Ready for agents
One API and CLI for people and autonomous workflows
FIG 0.3
Focused on incidents
Compact routine checks and preserve the failures that need attention
1.0 Monitor
Start in seconds
Add an endpoint, choose an interval, and Pulse handles the rest. Or have your agent manage.
2.0 Detect
From failure to recovery
Pulse traces each check from first failure through recovery. Status, incidents, and alerts in sync.
Live Monitor
Inference
Up
GET
inference.superposition.app/health
42 ms
24 hours agoNow
Up
Verifying
Down
Verifying
Recovered
3.0 Respond
Automatic history
Pulse confirms outages, records each transition, and sends clear alerts. Add reports as you fix.
Incident
Inference availability
- Started
- —
- Duration
- —
- Cause
- —
No incidents yet
Possible dependency overlap
Vercel Runtime
View Provider Incident
Provider incident began 3 minutes before this outage
Your services are operational
Pulse will send one clear alert when the state changes.
4.0 Attribute
Is it us,
or them?
Add OpenAI, Stripe, or 20+ services in one click. Incidents appear beneath your own services.
Add Dependency
One click, no credentials
Dependencies
Beside your own checks
Pulse mirrors each provider's published status in real time
Works with the services you already run on
5.0 Publish
Status users trust
Every deployment serves a public status page from live monitor state. Nothing extra to run.
System Status
Last updated 12:04:52 UTC
Inference
Ongoing since 12:04 UTC · 42s elapsed
Connection timeout
Production
View GroupRecent Incidents
Powered by Pulse
6.0 Automate
Human first, agent first
Run pulsectl yourself, or hand your agent the CLI and let it manage setup and maintenance.
$
7.0 Operate
Built to stay small
Pulse compacts routine checks and preserves history within a minimal storage budget.
Database Health
Healthy
Storage remains within its configured budget
118 MBof 500 MB
24%
Rollups62 MB
Exceptions18 MB
Incidents9 MB
Recent checks3 MB
Core and indexes22 MB
- Mode
- Full detail
- Recent checks
- 48 hours
- Hourly rollups
- 30 days
- Scheduler coverage
- 99.99%
- Next action
- None required
8.0 Architecture
Vercel native
Pulse is Vercel-native, Postgres-backed, and small enough to audit or extend.
Hover a system for its role. Click to zoom in.
01Minimal design
No queues, no workers, no servers to babysit
02Exception-first history
Keep detail where it carries meaning
03One control plane
Serve people and agents consistently
9.0 Harden
Trust nothing
Every target, response, redirect, and DNS answer is treated as hostile by design.
SSRF-proof checking
Only publicly routable addresses. Every DNS answer is validated and the socket is pinned to it.
Hardened auth
Argon2id passwords, gated first-admin bootstrap, rate-limited login, hashed tokens with lineage.
Hostile-proof CLI
pulsectl
treats server responses as untrusted and rejects control
characters outright.
lib · checker
enforcing
$ check https://inference.superposition.app/health resolve inference.superposition.app → 172.64.3.14 policy public range · not private, not link-local pin socket pinned to validated address tls TLS 1.3 · certificate valid result 200 OK in 42 ms → UP
Blocked
2 attempts
✗
redirect → 169.254.169.254
link-local · BLOCKED_TARGET
✗
target resolves to 10.0.0.8
private range · refused
Untrusted providers
Status feeds read allowlisted hosts only. No cross-host redirects, 512 KB caps, XML entities off.
Locked-down browser
Strict CSP, frame-ancestors none, nosniff, and preloaded HSTS on every response.
Abuse containment
Rate limits without amplification. Scheduler leases and dedup make duplicate alerts impossible.
Ship it locked down. Read the hardening guide
10.0 Accelerate
Fast by doing less
Fewer renders, fewer fetches, fewer statements. First paint to cron tick, every layer does less.
01Runtime
Shells stream while data loads. Revisits render from cache and rows prefetch on hover.
shell600 ms →
0
shell painted0 ms
rollup read · 24h0.078 ms
hoverprefetched
revisitcache hit
paint first, fetch later
02Scale
Cost follows change, not fleet size. More monitors do not mean more round trips.
cron tick401 →
5
read configedge config
registry sync5 stmts
alert fan-out1 insert
outbox claim0.627 ms
tick completesteady state
same cost at any fleet size
03Database
Set-based SQL and rollup reads. One statement does what four hundred used to.
queries0.082
ms median
uptime 24h4.894 ms
status rollups0.684 ms
outbox claim0.627 ms
detail rollups 30d0.288 ms
detail rollups 7d0.122 ms
incidents list0.119 ms
palette lookup0.108 ms
+12 more · none over 5 ms
One prompt · Live in 5 minutes
Deploy monitoring
you control
Codex discovers every app you run and monitors them all.
Copies the full agent deploy prompt