AI Tools & Discussions in Agent Harness

Infrastructure, orchestrators, and task runners that wrap around LLM coding agents — covering session management, context delivery, worktree isolation, architecture enforcement, and issue-to-PR pipelines.

Agent Harness Tools (316)

foreman harness for Grok Bot crews

Grok Bot Crew CLI Harness

Agent Harness Multi-agent Systems Observability

Cindy

Open Source Local AI Agent

Agent Harness Multi-agent Systems AI Coding Asst.

TokenEconomizer

Coding Agent Cost Reduction Plugin

AI Coding Asst.Compute Opt.Agent Harness

Kubic

Open Source Terminal Coding Agent

AI Coding Asst.CLI Assistants Agent Harness

MausBot

Open Source AI Agent Chat App

Multi-agent Systems Agent Harness Workflow Automation

Mouse-1

Agent Decision Model API

Decision Models Agent Harness AI Infrastructure

Nitely

Multi Repo Coding Agent Orchestrator

Multi-agent Systems Agent Harness AI Coding Asst.

SuperQode

Open Source Coding Agent Harness

Agent Harness AI Coding Asst.Multi-agent Systems

Distill

Open Source Terminal Coding Agent

Agent Harness AI Coding Asst.CLI Assistants

nightmux

Tmux Control for AI Coding Agents

Agent Harness Terminal Mgmt Task Automation

Agent Harness Discussions

Start a new discussion about Agent Harness

Theo Park

October 9, 2026·Weekend builder exploring tiny…

Claude Code Mods expose the agent loop to plugin code

Claude Code now lets you write JS/TS that changes how the agent works. Prompts, tool arguments, retries, results, commands, shared state, even the terminal UI. These are called Mods, shipped inside plugins, available in CLI v2.1.287. Lots of room to build workflows around how you actually work. Over…

Theo Park

October 7, 2026·Weekend builder exploring tiny…

What happens if the mod guarding a tool call crashes?

One detail in the Claude Mods docs caught my eye: if a hook throws before calling the next handler, Claude Code normally skips it and continues. A guard can instead handle the error and deny the call. That makes the error path part of the permission policy. Curious whether mod authors will show that…

Mateo Rivera

October 7, 2026·Open-source tinkerer exploring local…

Reflection announces Beam with an open-weight release planned for October

October 5. Reflection announced Beam, a sparse mixture-of-experts model aimed at coding, reasoning and agentic tasks. It has 501B total parameters and activates 23B per token. Reflection describes performance as competitive with GLM-5.2 and approaching Qwen3.8-Max; these are the developer’s comparis…

Priya Shah

October 7, 2026·Product designer learning to…

Does 23B active tell us much about hosting Beam?

Beam’s 23B active parameters sound appealing, but the model has 501B parameters overall. The active count alone doesn’t tell me what a usable deployment looks like. Once the weights land, I’d like to see a concrete hosting setup and a long coding task on it. What would you want to know first: memory…