pre-mvp

GitHub for data

Branches work like git. On data.

Fork, commit, diff, merge — on a graph. Every commit stays queryable.

Software solved collaboration with branches and history. Legra gives data the same workflow: people and agents work on forks, hand them on, and merge when the rules are satisfied.

Branch

Fork production, or any branch, in constant time. Copy-on-write: the fork shares every unchanged page.

Commit

Immutable, signed, content-addressed. A commit names the graph state and who made it.

Diff

Quads added, quads removed, between any two commits or branches.

Merge

Three-way merge on the graph. Conflicts are named; history keeps both parents.

Full history

Ask the graph what it knew last March.

Every commit stays addressable. A query against any past state costs what a cold query costs today: no snapshots, no restores, no second copy.

Same question · three commits
Commit 3f9a · today
Current answer
Commit 8c21 · last quarter
What the graph knew then
Commit 1e07 · the audit date
What the auditor saw

Content-addressed pages are shared across commits, so history costs storage only where the graph actually changed.