AutosClaw – security first *claw with live chat to any agent session
Claude orchestration with live dashboards and agent-spawning—well-built but competes with Anthropic, OpenAI infrastructure.
🔁 REPL for managing AI agents across multiple projects, round-robin
Orchestrates Claude across projects, but it's a thin wrapper around Claude Code CLI.
Engineers managing multiple projects with Claude Code simultaneously
Continue IDE · Cursor · GitHub Copilot Chat
It works like this:
1. Put a `roundsman.json` in any folder you work on with Claude Code
2. Run `roundsman` from anywhere on your machine
3. Roundsman will then start visiting all your projects in order, in round-robin fashion. Upon each visit, you are expected to give a prompt. Roundsman will send that prompt to Claude, who will do the actual work. While Claude is working, Roundsman will take you to the next available project, and so-on. If all Claudes are working, you will wait until one of the projects needs you. You'll see Claude's outputs while you wait.
It is kept very, stupidly, simple on purpose. I do not like windows/panels/tabs, I do not like "TUI" theatrics, and I do not like tools that I have to remember more than 2 things to use. So I made this.
(There are a few advanced features available too - such as slash-commands, obviously. Like `/snooze 13` (to stop visiting a given project for 13 minutes), or `/drop` (to remove a project from the round robin list), or even `/loop 88 fix all the bugs` which will make it tell Claude to `"fix all the bugs"` 88 times. See the README for more.)
Enjoy!
Claude orchestration with live dashboards and agent-spawning—well-built but competes with Anthropic, OpenAI infrastructure.
Auto-spawning fresh agents at 50% context beats manual session restarts.
This is not just a chat wrapper — the app ships agent-first features (permission cards, plan mode, thinking blocks, subagent visualization) plus developer niceties like forked sessions, cost/token tracking and 1M-context support. The terminal experience looks technically ambitious too: multi-tab Claude Code terminals with xterm.js + WebGL rendering and per-tab chat/terminal modes. It’s a focused, feature-heavy tool for people already committed to Claude ecosystems; casual users won't care, but if you run agents locally it's a very compelling desktop workflow.
Auto-detects Claude Code sessions and streams agent decisions as interactive node graphs.
OpenClaw wrapper with 12-agent swarm, but MVP-stage claims outpace deliverables and no independent verification.
Claude Code orchestrator with SSE status detection, PWA dashboard, and tmux isolation—genuinely useful for multi-agent work.