Schema changes are where databases get hurt. A migration drops a column someone still used. A type change locks a big table at the worst moment. A tool decides to “tidy up” and removes something nobody asked it to touch. And after it’s all done, nobody is quite sure the database matches what the team thinks it looks like.
ForgeSchema is a PostgreSQL schema evolution engine written in Go, built around four steps: inspect, plan, apply and verify. It explains every change before it runs and drops nothing you didn’t list. You describe the schema you want, review the plan it writes, apply it in a transaction, and then check the database with a separate read-only process that doesn’t trust the apply step.
The Pitch in Four Numbers
- 4 commands: inspect, plan, apply and verify.
- 3 PostgreSQL versions supported: 16, 17 and 18.
- 0 drops you didn’t list. Nothing disappears unless the plan says so and you approved it.
- 1 transaction per apply, with indexes built online.
How It Works
Inspect reads the live schema: defaults, identity columns, unique, check and foreign key constraints, enums, arrays and views. Plan compares it with the schema you want and writes an explicit, reviewable plan, including drops and type changes when you ask for them, with an explanation of each step. Plans can be posted as comments on pull requests, so schema changes get reviewed like code.
Apply runs the reviewed plan in a transaction, building indexes online so tables aren’t locked longer than they need to be. Only supported, reviewed plans can be applied. Verify then checks whether the database structurally matches the desired schema or has drifted, without changing anything. Because verify is independent of apply, it’s useful on its own: run it in CI against staging and production and you’ll know when someone made a change outside the process.
The desired schema lives in files in your repository, so development, staging and production can all be checked against the same source. A reproducible walkthrough on a disposable database shows the whole cycle, and a capability matrix in the repository spells out exactly what’s supported.
Who It’s For
- Teams on PostgreSQL who want to see every schema change before it runs.
- Reviewers who want migration plans in the pull request, not hidden in a script.
- Platform and DBA teams checking that staging and production haven’t drifted.
- Anyone who’s been burned by a migration tool dropping something it shouldn’t.
Limits
ForgeSchema is PostgreSQL only, and still in development. Only features in its capability matrix can be applied, and anything else is refused instead of guessed at. The project publishes scope, not performance figures.
The Four-Question Screening
Every project featured here gets the same four questions, the way you’d screen a startup idea.
1. Where is the money in it?
The engine is open source under Apache 2.0, with no paid tier. Database change management is a real market: teams pay for tools that make schema changes safe and auditable, especially in regulated industries. The paid paths are a hosted review and approval workflow with audit history, drift monitoring across many databases, and support for large teams. Plans that explain themselves in a pull request are a natural fit for compliance-minded companies.
2. What real problem does it solve?
Schema changes nobody fully understood before they ran, and databases nobody can prove are in the expected shape. Both lead to outages and slow, nervous releases.
3. The technology in plain words
Think of renovating a house with a careful builder. First they survey what’s there. Then they hand you a written plan of every wall they’ll move, and they won’t knock anything down that isn’t on the list. They do the work in one go. Then a separate inspector, who didn’t do the work, walks through and checks the house matches the drawings.
4. Moat and differentiation
Migration tools are plentiful, and many either run hand-written scripts or generate changes with little explanation. ForgeSchema’s edge is explanation and restraint: every step explained in advance, nothing dropped that you didn’t list, and verification that doesn’t trust the apply step. The moat that can grow is depth on PostgreSQL’s harder features and its place in the pull request workflow, which makes it part of how a team ships.
Why It Matters
Most database outages start with a change. A tool that makes every change visible, reviewable and checkable lowers the risk of the riskiest moment in a database’s life.
Follow the Project
Read the introduction, learn about the engine and the schema format, try the demo, see the roadmap, or get the code from the ForgeSchema repository on GitHub. ForgeSchema is listed with the other open-source projects on SoftDevStudio.com, and more featured projects are in the Projects section.