Cast AI analyzes every PostgreSQL and MySQL query in real time, with no sampling and read-only access, flags the slow ones, and recommends the index or rewrite to ship. When you’re ready, route traffic through the proxy to cache repeated reads and pool connections.
480 → 40customer_idWrites and uncached reads pass through to your database.
Trusted by 2,100+ companies
Slow queries, missing indexes, and connection spikes tend to surface as incidents. Native tools record what your queries do, but turning that raw data into the right fix takes expertise.
Cache hit rate with Database Optimizer.
Read the storyCast AI analyzes every query in real time, with no sampling, so you find slow queries and regressions before users report them.
Cast AI recommends the exact index or rewrite to ship, ranked by the database time it would reclaim.
Serve repeated reads from cache, with no application code changes.
Cast AI multiplexes many application connections onto a smaller pool of database connections, so autoscaling spikes don’t exhaust your limit.
“Cost optimization might be nice, but I see big value in this tooling helping us with our database administration work. It replaces a lot of what a dedicated DBA would do, automatically, in production, without any manual tuning.”
A lightweight agent with read-only credentials analyzes every query in real time and recommends indexes and rewrites, with the first in 24 to 48 hours.
Run the proxy in Passthrough mode to see projected cache hit rates before anything is cached. Each recommendation shows before and after plans.
Switch to Auto mode and DBO decides what to cache, or set rules yourself in Manual mode.
Keep PostgreSQL or MySQL, whether you use Amazon RDS, Google Cloud SQL, or a self-hosted database on AWS, Google Cloud, or Azure. Connect through a connection string change, with no migrations or code changes.
Connect an instance to see latency, slow queries and connections in one view.
Connect a read-only IAM role on AWS or GCP and DBO finds your database instances, or add a single instance by hand.
DBO ranks instances by optimization opportunity and points to the best one to start with, with projected gains.
The proxy runs in your Kubernetes cluster, and MySQL deployments also support ECS.
Cached data stays in memory in your cluster, and query patterns are analyzed there.
Read how it works, see what other teams got, or talk to a specialist about your setup.
Architecture, setup guides, and supported environments.
Read the Database Optimizer documentationResults from teams running Cast AI in production, from lower costs to engineering time saved.
See customer resultsWalk through your setup with an engineer and see where Cast AI fits.
Book a discovery callCan’t find what you’re looking for?
Contact usIt’s part of the Cast AI platform. A read-only Performance Advisor recommends indexes and query rewrites for PostgreSQL and MySQL, and a proxy caches repeated reads and pools connections.
PostgreSQL’s pg_stat_statements and MySQL’s performance_schema give you raw query data that takes expertise to interpret. The Performance Advisor does that analysis for you and ranks its SQL recommendations by database time reclaimed. If you build your own cache, you also own invalidation. Every write that passes through the proxy triggers Smart Invalidation, which parses the write’s SQL and evicts exactly the affected entries. It needs no WAL access, so it works on managed databases like Amazon RDS and Google Cloud SQL.
The Performance Advisor connects with read-only credentials and never changes your schema. You review and run any recommendation yourself. The proxy only sits in the path once you route traffic through it, and it starts in Passthrough mode, where nothing is cached.
DBO runs in your own infrastructure, installed with Helm or Terraform: the Performance Advisor as a lightweight agent in your Kubernetes cluster, and the proxy as a deployment you reach by changing your connection string. For PostgreSQL, enable pg_stat_statements first, which is a parameter group setting on RDS, a database flag on Cloud SQL, or shared_preload_libraries when self-hosted. MySQL deployments also support ECS.
DBO never applies index recommendations for you. Each one comes with the exact SQL, a plain-English explanation, and a before and after plan comparison showing the projected impact. For caching, Passthrough mode shows projected cache hit rates before anything is cached.
Cached data stays in memory in your cluster, and query patterns are analyzed there. For query rewrites, the Performance Advisor sends only an anonymized query template to an LLM, with values and identifiers replaced by placeholders. No query values, results, or customer identifiers leave your cluster.
For PostgreSQL, pooling runs in transaction mode, so session-level features don’t work through the proxy, including LISTEN/NOTIFY and session-level SET commands. Applications that use server-side prepared statements, which includes some Rails, Django, and Hibernate setups, need them disabled. Route connections that need these features around the pool.
Run it as a deployment with at least two replicas and a PodDisruptionBudget. Connections drain gracefully during planned restarts, and you can fall back to the database by reverting the connection string host, with no code changes.
#1 on G2262 reviewsStart with a read-only look at your queries. Turn on caching when you’re ready.
