Early access

Real data.
Zero risk.

Glass creates an instant, isolated sandbox for your PostgreSQL database called a frame. Spin up as many as you need in seconds. Test migrations, run AI agents, reproduce bugs. Production stays untouched.

terminal
$ glass frame create my-migration
# Frame created in 4.2s
postgresql://glass:tok_x9k2@localhost:5544/my-migration
$ psql $FRAME_URL
# Full production data. Fully isolated.
$ glass frame delete my-migration
Frame deleted. Production untouched.

Databases are terrifying
to experiment with

You need to test a migration, try a risky query, or let an AI agent loose on your schema. But your database has real data in it. So you end up with bad workarounds.

AI agents need a real sandbox

Claude Code, Cursor, Copilot are writing migrations now. Pointing them at production is reckless. Seeds mean they can't validate their work.

Seed scripts lie

They never quite match production. You fix the bug locally, push it, and it breaks in prod because the data was different.

Shared staging is chaos

Two developers, one staging database. Conflicting migrations, dirty state, and nobody trusts the data anyway.

From zero to isolated
database in minutes

01

Connect once

Run glass init with your connection string. Glass takes a one-time snapshot. This is the only slow step. Every sandbox after this is instant.

$ glass init postgresql://user:pass@host:5432/mydb
02

Create a frame

One command. Glass uses copy-on-write to spin up a fully isolated sandbox in seconds. No data movement whatsoever.

$ glass frame create my-migration
→ postgresql://glass:tok_x9k2@localhost:5544/my-migration
03

Use it anywhere

Standard Postgres connection string. psql, your ORM, your test suite, an AI agent. Anything that speaks Postgres just works. No Glass SDK.

$ psql $FRAME_URL
$ CLAUDE_DB_URL=$FRAME_URL claude code
04

Delete when done

One command cleans up everything. Production was never at risk. Run as many frames as you want, whenever you need them.

$ glass frame delete my-migration
✓ Done. Production untouched.

Built for the moments
that actually matter

🔀

Test database migrations

Run your ALTER TABLE on a frame, verify it works, then apply to production with confidence. If something breaks, delete the frame and try again.

🤖

Safe AI agent access

Hand a frame connection string to Claude Code or any AI agent. Real production data. Zero ability to affect production.

🐛

Reproduce production bugs

A bug against production data is hard to reproduce with seeds. Create a frame with the exact production state the bug occurred in.

🚀

Onboard new developers

Instead of half a day setting up a local database, one command gives a new team member a frame with real data, instantly.

🔬

Schema experimentation

Try a new index, foreign key, or restructured table freely. No consequences, no cleanup, no anxiety.

🔀

Parallel feature work

Two developers, two conflicting schema changes. Each gets their own frame. No stepping on each other's toes.

Instant after setup

Snapshot once. Every frame after that in under 10 seconds.

🔒

Complete isolation

Copy-on-write. Writes in a frame never reach production or other frames.

📦

Data stays with you

Glass runs in your own infrastructure. Your data never leaves your environment.

🔌

Standard Postgres

Any tool that speaks Postgres just works. No SDK, no driver, no integration work.

Get early access

Glass is in early development. Join the waitlist and we'll reach out when it's ready to use.

No spam. We'll only email you when Glass is ready.