Early development — pre-beta

Orchestrate teams of AI agents, not just one chat

DysonHarness is a multitasking orchestrator of smaller agents for large-scale work — not a single-agent chat wrapper. Delegate, parallelize, and keep every thread of the job moving at once.

View on GitHub

DysonHarness agent shell showing Work mode with subagent results and live tools

Multi-agent orchestration

Split big jobs across focused agents that work in parallel and report back — so large-scale work finishes as a team effort, not a single long conversation.

Plan → enhance → build → review → fix

Have the design written down and reviewed before any code changes happen, so bigger work lands the way you intended.

Persistent sessions

Sessions persist and resume, so long-running work survives restarts and picks up right where it left off.

Works with your provider

Connect to OpenAI-compatible providers and run DysonHarness with the capable models you already use.

Ready to put a team of agents on the job?

Free and open source under AGPL-3.0. Pre-beta — expect rough edges and rapid change.

Get started