gluerun
Orchestration engine for parallel AI coding agents
TLDR
SYNOPSIS
gluerun command [arguments] [options]
DESCRIPTION
gluerun is an orchestration engine that coordinates multiple AI coding agents to work in parallel on a software repository. It implements a multi-tier scheduling model with lease management, audit workflows, and isolated execution using git worktrees so that concurrent agents do not interfere with each other.Each task is dispatched to a supported runner CLI (such as claude or codex), driven through a pipeline, and then integrated back with an audit trail of the evidence produced. The engine is installed once per machine and pinned per repository through a .gluerun-version file, so improvements propagate by bumping the version rather than re-copying scripts.
PARAMETERS
init
Scaffold the initial configuration and documentation in a consumer repository.doctor
Validate required dependencies and the repository configuration.reconcile [--actuate]
Run one orchestration cycle: import, recover, integrate, dispatch, and snapshot. Without --actuate the cycle is a dry run.drive TASK-ID
Execute a single task through the agent pipeline.auto
Run the autonomy loop, dispatching work until a wall-clock budget is exhausted.update version
Pin a new engine version for the current repository.
CAVEATS
Requires Bash 4 or newer, Python 3, Git, and at least one supported agent runner CLI on the PATH. Because the engine grants autonomous agents the ability to edit and commit code, run it only on repositories where you can review the resulting changes.
HISTORY
gluerun (glueRun-go) is a community project written primarily in Bash and Python, built around git-worktree isolation and task evidence for agentic coding workflows. It is distributed under the GPL-3.0 license.
SEE ALSO
git-worktree(1), claude(1)
