Get started

From install to your first parallel run in three steps.

Meshloop installs as a single binary and works inside your existing Git repository and AI coding CLI sessions.

1 · Install

Install the binary & check readiness

Install from crates.io or source and verify your Git and agent CLI setup:

cargo install meshloop
meshloop doctor

2 · Plan & Review

Generate and approve the task graph

Use the slash commands in your coding session (or the CLI) to plan and approve:

/meshloop:plan
/meshloop:review-plan

3 · Run & Accept

Execute in sandboxes, then merge

Run tasks in parallel worktrees and merge only the verified result you accept:

/meshloop:run
/meshloop:accept

View full documentation & source on GitHub ↗ Read how isolation & repair work

Frequently asked questions

Common questions before you start.

Which AI coding tools work with Meshloop?

Meshloop works with Claude Code, OpenAI Codex CLI, Gemini CLI / Antigravity, and Grok CLI. You can assign different models or tools to planning, building, and reviewing.

Can a parallel worker accidentally overwrite my uncommitted local changes?

No. Every task executes in its own disposable Git worktree under .meshloop-worktrees/. Nothing touches your main checkout until you explicitly run /meshloop:accept.

Do I need AgentFlow SDLC to use Meshloop?

No. Meshloop is 100% standalone. If you also use AgentFlow SDLC for team-level role governance and release gates, the two systems complement each other cleanly.