Harness
Tiny apps for agents and humans
Start new thread
Matthew Kenney

Harness - Tiny apps for agents and humans

by•
Harness is standalone desktop apps that talk to any AI agent through localhost APIs. Not a plugin, not a platform. Just apps on your machine. The agent reads your kanban board, respects your "don't do" list, sends you alerts, leaves async questions, and checks how much time it has left. Works with Claude Code, Codex, Cursor, or anything that can make HTTP requests. Drop a markdown file in your project and the agent knows the apps exist. No accounts, no cloud, no tracking.

Add a comment

Replies

Best
Matthew Kenney
Maker
📌
Hey! I built Harness because I kept running into the same problem: I'm working with an AI agent, it finishes a task, and I have no idea what it did or what it needs from me. The terminal scrolls past and context is lost. I wanted simple, physical tools on my desktop. A board I can glance at, a sound when something breaks, a place the agent can leave me a note. Not another tab, not another SaaS dashboard. Every app exposes a localhost API. You drop a CLAUDE.md or AGENTS.md file in your project root and the agent just knows the tools are there. No config, no auth tokens, no setup wizard. The "don't" list in the Todo app is my favorite idea — you tell the agent what NOT to do, and it can see those guardrails but can't change them. Simple but surprisingly effective. Happy to answer questions about the architecture, the agent instructions format, or anything else. And if you have ideas for new apps in the suite, I'm listening.