Nothing had gone down yet
A classic monitor alerts at 95%, with no context and in the middle of the night. PulseDB sees the tablespace at 87% and growing 1.4 GB a day: it fills up on Thursday the 12th in the afternoon. There is no risk today.
Your database never sleeps. Neither does PulseDB.
Proactive AI monitoring for Oracle and PostgreSQL. It spots problems days before they arrive, explains the cause and leaves the decision in your hands.
From noise to signal.
Every night your database emits thousands of signals and almost all of them are noise. In 30 seconds, this is how PulseDB finds the one that matters.
It sees the problem before it exists.
PulseDB doesn't wait for something to reach 95%. It looks at the trend of the last 14 days and tells you when it will fill up, why and what options you have.
- 24/7
- Continuous monitoring of every instance
- 14 days
- Of trend, not just a threshold
- 1%
- Of one CPU as the load ceiling
Built so you stay in command.
Read-only, by design
The agents have no tool capable of writing to your databases.
ALTERNever changes structuresKILLNever kills sessionsRESTARTNever restarts anythingSETNever touches parameters
No agents
Nothing is installed on the database server. Oracle is queried through ORDS and PostgreSQL with a read-only role.
Without reading your data
Only the dictionary and statistics. Query literals are replaced with bind variables before reaching the AI model.
You decide. Always. PulseDB watches, analyzes and recommends, but no action happens without a DBA approving it. The only write is the one you enable in diagnostic mode, and only on your own account.
The alert that doesn't wake you.
Four levels. Alerts are grouped per database: if an instance goes down, you get one message, not forty.
Informational
Dashboard only. Nobody gets notified.
DashboardWarning
Dashboard and email during business hours. Repeats every two hours and never goes out by WhatsApp.
Dashboard · EmailSevere
WhatsApp within five minutes, at any hour. Repeats every 30 minutes and escalates to a backup after four hours.
Dashboard · Email · WhatsAppCritical
All channels immediately, every five minutes until someone acknowledges it. Only a declared maintenance window silences it.
All channelsEverything you need, in the first paragraph.
Every ticket arrives with a probable cause, evidence, and options with their risk and duration. When it is closed, PulseDB remembers what happened; twice a year, it projects your capacity twelve months ahead.
Confidence 0.86
Probable cause
users_04.dbf was left at a fixed size after the May 14 migration, and the daily growth of MOV_CAJA rose by 40%.
Evidence
- dba_data_files
- dba_free_space
- usage series · 14 days
- parameter change · 05/14
- previous incident #0761
Options
- Enable autoextend with a 32 GB cap
- Add a new 32 GB datafile
- Archive MOV_CAJA data older than 2024 (recovers 41 GB)
It remembers everything for you
When a ticket is closed it compares the initial state with the current one and proposes the text of the corrective action for you to confirm. If a problem repeats three times in 90 days, it proposes a new check that spends 24 hours in a lab before touching production.
ORCLPRD1 · total storageHigh risk
1598GB used of 2000
0.981fit r²
8–11months left
The spec sheet, no frills.
59checks: 33 Oracle, 25 PostgreSQL and 1 shared
36seconds per hour as the load ceiling per database
17seconds per hour measured in the Oracle pilot
0writes to your databases
- Engines
- Oracle 19c and PostgreSQL 16, verified against real databases. Aurora PostgreSQL connector written, not yet tested on a real cluster.
- Access
- Pre-approved ORDS REST handlers on Oracle; a read-only role on PostgreSQL.
- Users
- SELECT_CATALOG_ROLE and AUDIT_VIEWER on Oracle; pg_monitor on PostgreSQL. Automated tests try to write with that user and must fail.
- Load
- At most 36 seconds of database time per hour (≈1% of one CPU), watched by PulseDB itself.
- Diagnostics
- You enable it, for a limited window. ASH and AWR only if your Diagnostics Pack license allows it.
- AI model
- Via API (today Qwen, DeepSeek and Kimi depending on the task) or an OpenAI-API-compatible model on your own infrastructure. The data in your tables never leaves, under any circumstance.
- Audit
- Every action, yours or the agents', is recorded with the before and after state in a hash-chained log.
Choose where PulseDB lives.
The same product in two models: on a server we manage, or inside your own infrastructure.
External VPS
We host it
USD3,000
One-time initial payment: includes installation, configuration and go-live.
- Dedicated virtual server, managed by INNOVA
- Connection of your Oracle and PostgreSQL instances
- Dashboard, alerts and tickets from day one
- Subsequent monthly operating and administration fee: monitoring, maintenance, updates and support. Its amount is set in the proposal according to the number of databases.
Secure payment on the PayPal page. Accepts credit and debit cards, with or without a PayPal account.
Local VPS
On your infrastructure
Custom quote
For organizations that need everything to stay inside their network.
- Installation on your servers or private cloud
- Option of an AI model hosted by you
- Support from the INNOVA team
- Scope and price based on your architecture
Handled directly by Edmundo, from the INNOVA team.
Amounts in US dollars (USD). Before paying you will receive a proposal with the scope, the monthly fee and any applicable taxes. By paying you accept the Terms and conditions (in Spanish).
Let's start with just one database.
A demo on one of your instances, on your infrastructure, that ends with a risk inventory even if you don't hire us. All we need is a read-only user and an hour of your time.