HAR: an isolation harness for running multiple coding agents
HAR is an Apache 2.0 open-source agent harness that gives each task its own worktree and runs launch, verification, and teardown via CLI or MCP, with a local dashboard.
Running several coding agents shouldn’t turn your repo into a shared-state mess. HAR is an open-source agent harness for AI coding workflows, for developers who want repeatable isolation and verification. It helps you run one agent or a fleet with more control by giving each task an isolated worktree and running launch, verification, and teardown through a machine-readable harness contract. Key features: • CLI + MCP access – use the core harness from the command line or as an MCP server • Per-agent worktrees – each slot gets its own branch, ports, and, where needed, database • Deterministic verification – runs the project’s own checks and records a consistent result • Evidence trail – keeps logs, artifacts, and a validated tree hash for review • Local Mission Control – track repositories, worktrees, runs, validations, and artifacts in one dashboard It’s open-source (Apache License 2.0). Link in the reply 👇