PGRun
PGRun gives every coding agent, pull request, and CI job its own isolated, disposable, production-like PostgreSQL branch. It connects to an existing production database, creates a human-approved masked Safe Copy, and serves ordinary Postgres branches without moving or writing to production.
At a Glance
- Teams using coding agents such as Claude Code, Codex, or Cursor
- Software engineering and platform teams
- Organizations running GitHub Actions or GitLab CI
- Teams with production PostgreSQL on RDS, Aurora, Cloud SQL, Supabase, Neon, or self-hosted infrastructure
- +1 more
AI Tools by PGRun
(1)
PGX
Postgres Runtime for Disposable Databases
Discussions
No discussions yet
Be the first to start a discussion about PGRun
Latest News
PGX v0.1.0-beta.1 released as the first public beta
PGRun CI v1.1.0 released with framework-agnostic disposable PostgreSQL action for any stack, agent, or CI
PGRun CI v1.0.0 released for production-like and fast parallel PostgreSQL testing
pgrun-cli v0.4.0 released with templates, cancellation cleanup, timing output, and corrected signal exit codes
Products & Services
Managed service that connects to RDS, Aurora, Cloud SQL, Supabase, Neon, or self-hosted Postgres; builds a masked Safe Copy and creates isolated copy-on-write branches in about three seconds. Each branch provides a normal PostgreSQL connection URL and can be deleted or given a TTL.
Open-source Rust database runtime behind PGRun branches. It runs many isolated, short-lived Postgres-compatible databases in one process, using shared immutable state, copy-on-write storage, mint-on-connect, automatic cleanup, and resource limits.
Static Go CLI and MCP server for authentication, production-source setup, protection review, Safe Copy creation, branch create/list/get/delete/exec, templates, JSON output, TTLs, and agent skill installation.
GitHub Actions integration that provisions a disposable database per job or parallel worker, from sanitized production-like data or a schema-only template, exposes DATABASE_URL, runs with any stack, and cleans up on success, failure, cancellation, or TTL expiry.
Market Position
PGRun positions itself as a hosted, agent-first alternative to shared staging and dump/restore workflows: it keeps production where it is, masks data before copying, and offers fast copy-on-write branches with ordinary PostgreSQL connectivity. Its researched alternatives include Neon, Xata, Layerbase Cloud/local branching, and self-hosted DBLab Engine; PGRun differentiates on external-production Safe Copies, pre-copy masking, agent/CI automation, and no required migration to a new Postgres host.
Leadership
Founders
Alex Shapalov
Founder of PGRun; describes himself publicly as a Postgres and open-source developer and also as the creator of JobsGPT and PgBot. GitHub repositories under pgrundev show him as the principal maintainer/releaser of PGX, pgrun-cli, and PGRun CI.
Executive Team
Alex Shapalov
Founder
Postgres and open-source developer; public profiles identify him as Founder of PGRun and creator/maintainer of related Postgres projects including PgBot.
Founding Story
PGRun was started around the problem that multiple coding agents sharing one staging database cause migration collisions, shared state, and untrustworthy test results. Its initial vision was to let agents branch a protected production-like Postgres database on demand, run migrations and tests in isolation, then discard the branch.
Business Model
Revenue Model
Usage-based hosted Postgres: branch compute is billed by branch-hour and storage by GB-month; managed dedicated Postgres is billed per database per month plus storage. Branches are free during beta with included credit.
Pricing Tiers
$10 of credit included; no credit card required. Beta limits include up to 10 live branches per project and databases up to 10 GB.
Billed by the second while a branch is running.
Hourly measurement of the Safe Copy plus data written by copy-on-write branches.
1 vCPU, 2 GB RAM, plus $0.10/GB-month storage.
2 vCPU, 4 GB RAM, plus $0.10/GB-month storage.
2 vCPU, 8 GB RAM, plus $0.10/GB-month storage.
4 vCPU, 8 GB RAM, plus $0.10/GB-month storage.
6 vCPU, 12 GB RAM, plus $0.10/GB-month storage.
8 vCPU, 24 GB RAM, plus $0.10/GB-month storage.
Target Markets
- Teams using coding agents such as Claude Code, Codex, or Cursor
- Software engineering and platform teams
- Organizations running GitHub Actions or GitLab CI
- Teams with production PostgreSQL on RDS, Aurora, Cloud SQL, Supabase, Neon, or self-hosted infrastructure
- Rails, Django, Prisma, Drizzle, Go, and other Postgres application teams
- Coding-agent database sandboxes
- Per-pull-request migrations and preview environments
- Production-like CI and migration safety checks
- Parallel test workers with isolated databases
- Short-lived development and disposable test databases
- Testing schema changes against masked production data