01 / FORK
A fresh starting point.
Connect a Postgres source and create an isolated copy for the task at hand. Every experiment gets its own space.
YOUR SOURCE → YOUR FORK A NEW BRANCH OF POSSIBILITYPOSTGRES / EARLY DEVELOPMENT
An AI workspace for your database.
Give your agents an isolated copy, a real task, and room to figure it out. Then see exactly what changed.
your databaseSOURCE
room to experimentFORK
ONE IDEA. ITS OWN WORKSPACE.
Let your agents
experiment.
YOUR DATA. A FRESH PERSPECTIVE.
TAKE A LOOK
01 THE WORKSPACE
A question. A fork.
A change you can see.
Move from “what if” to “here’s what happened.”
Your agent does the work in a fork.
You stay close to the outcome.
Forkdenplaygroundcountry-backfillFORK
Ready to exploreFORK OVERVIEW
Every change, in view.
Query activityILLUSTRATIVE SESSION
ReadWriteCheckAll inside your fork
Rows affected128
Data checks3 / 3
Source databaseSeparate
PROPOSED SQLPOSTGRESQL
UPDATE users AS u
SET country = p.country
FROM profiles AS p
WHERE u.id = p.user_id
AND u.country IS NULL
AND p.country IS NOT NULL
AND u.tenant_id = 'demo';CHANGE PREVIEW4 SAMPLE ROWS
Run the example to inspect the changes in the fork.
YOUR DATA RULESNOT RUN YET
Go ahead. Try a task and inspect the result.A PREVIEW OF WHAT WE’RE BUILDING
02 HOW IT WORKS
A little space.
A lot of possibility.
The planned workflow is simple:
make space, let your agent work,
then decide what comes next.
02 / EXPERIMENT
Bring your own agent.
Ask a question, try a backfill, or rehearse a migration. MCP and CLI are the planned ways into the workspace.
MCP + CLI · PLANNED03 / REVIEW
See the whole change.
Review the SQL, affected rows, and your data checks. Export the work you want to take forward, then close the fork.
INSPECT → EXPORT → CLOSE03 ROOM FOR THE REAL WORK
What’s on
your mind?
Start with a question.
Leave with something you can inspect.
04Follow your curiosity.Explore your schema, investigate a question, and try a better query.
04 FORKDEN REPLAY PLANNED OSS
One more way
to check your work.
Run the same query set against a baseline and a changed fork. Compare new errors, data checks, and measured timings.
Planned as a standalone local tool, and a natural part of the Forkden workspace.
A little more about Replayreplay / comparisonCONCEPT PREVIEW
query-set.sql × schema-change.sql
Same queries.
A clearer picture.
A clearer picture.
BASELINE48 / 48queries passed
CANDIDATE47 / 48queries passed
1 new SQL errorcolumn “legacy_status” does not exist
42703Unrelated tenant data unchangedPASSED
A FEW THINGS TO KNOW
Good questions.
Can an agent change my source database?
The workspace is being designed around isolated forks. Agent operations happen in the fork; the initial scope does not include automatically applying changes to your source. Exported SQL goes through your own review and release workflow.
Which databases are you starting with?
PostgreSQL is the proposed first scope. We’re shaping the connection and copying workflow around the first real tasks teams want to do.
Can I use my own AI agent?
That’s the direction. MCP and CLI are the planned entry points, so you can bring your existing agent into a database workspace instead of changing how you work.
Where does Replay fit in?
Forkden is the AI database workspace. Replay is a planned supporting feature and standalone open source tool for comparing a query set before and after a change. It follows the core fork, agent, and review workflow.
THE NEXT CHAPTERIN DEVELOPMENT
Make room for
your next idea.
Forkden is taking shape.
Follow the build and help shape
what comes next.