GitButler

Key features

• Virtual branches Work on multiple branches in one working directory at the same time, so parallel lines of work no longer fight over a single checkout.[1][2]

• Hunk-level staging Sort changes onto branches as they are made, staging individual hunks or dragging changes onto branches and commits.[1]

• Stacked pull requests Push stacked branches to GitHub, GitLab, or Bitbucket as pull requests, keeping dependent changes reviewable in order.[1][2]

• Companion but CLI Run status, commit, push, and conflict-resolution commands from the terminal with the same workspace model as the desktop client.[1][2]

• MCP integration Expose GitButler workspace operations to coding agents through the Model Context Protocol, including rules support.[1][2]

• AI commit messages Generate commit messages from cloud models or from local Ollama and LM Studio, keeping message drafting inside your own machine when preferred.[1][2]

• Undo timeline Restore file states from past snapshots, recovering work that staging mistakes or bad rebases would otherwise lose.[1]

• Full Git compatibility Open any existing repository and keep using signing, LFS, submodules, and worktrees with the forge hosting you already have.[1][2]

Who will love it

• Developers juggling parallel work Virtual branches with hunk-level sorting match the reality of interrupted work, context switches, and half-finished ideas sharing one tree.[1][2]

• Teams landing stacked pull requests Pushing ordered stacks to GitHub, GitLab, or Bitbucket fits teams that review dependent changes as a sequence instead of one giant diff.[1][2]

• Agent-assisted developers The MCP integration and but CLI give terminals and coding agents the same workspace operations as the desktop client.[1][2]

Who should avoid it

• Fair Source license, not OSI-approved The source is available under the Functional Source License 1.1 with MIT conversion after two years, which is not an OSI-approved open-source license.[1][2][3]

• No documented enterprise support No support tier or SLA was found on official pages during this review, so procurement-driven teams cannot complete due diligence from public sources.[1]

• Overkill for single-branch work If your workflow is one branch at a time in plain Git, the virtual-branch model adds concepts without a matching payoff.[1][2]

• What the reports suggest No community sample was inspected, so no sentiment analysis is possible. The assessment above is documentation-based until firsthand reports are collected.[1][2]

Alternatives and compatibility