AI-Native
Version Control
An experimental version control system for humans and AI agents working on the same code.
Owned symbols, sealed stitches, and the why kept with every change.
macOS and Linux
$ curl -fsSL https://weave-cli.com/install.sh | shWindows (PowerShell)
$ irm https://weave-cli.com/install.ps1 | iexThen weave login and weave init to connect your project. Direct downloads and details
Symbols, Not Lines
The unit of change is the symbol: a function, class, method or import. Your private draft is sealed into a stitch, the weave check publishes it, and weaved means deployable for that workstream.
Owners, Not Clashes
Every symbol has one owner. Edits to someone else's symbols become suggestions they accept or reject, and a symbol changed differently in two workstreams becomes a change request carrying both sides' why.
Decisions for Agents
Record why the code is the way it is with weave stitch --why. Decision records are anchored to symbols, versioned with the code, and served to AI agents over MCP in a small token budget.
$ weave workstream new feature-greet
Created workstream 'feature-greet' from 'main' (ownership repo, inherit auto).
Switched to workstream 'feature-greet'.
$ weave add .
Added to your draft in 'feature-greet':
edit app.py::greet (function)
$ weave stitch -i "Friendlier greeting" -w "support asked for it" -y
Decision D-e437 recorded with this stitch: Friendlier greeting
Weaved 6f1e308815 into 'feature-greet'.
$ weave promote 6f1e308815 --to main
Promoted 6f1e308815 to 'main': weaved as 6f1e308815.