simonstaton

simonstaton / AgentManager

Public

Conduct autonomous Claude Code agents at scale safely. Human-on-the-loop orchestration platform with kill switch, cost tracking, and inter-agent messaging.

agent-framework agentic agents ai ai-agents
14
2
100% credibility
Found Feb 22, 2026 at 13 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

AgentManager is a platform for orchestrating teams of AI coding agents with real-time monitoring, inter-agent communication, and built-in safety features like a multi-layer kill switch.

How It Works

1
🔍 Discover AgentManager

You find a tool that lets AI agents work together like a development team, safely and with easy controls.

2
⚙️ Set it up quickly

Follow simple steps to run it on your computer or cloud, connecting your preferred AI service.

3
🔑 Log in securely

Enter a private password to access the dashboard where you manage your agents.

4
🚀 Launch your first agent

Give an instruction and watch your AI agent spring to life, ready to code or analyze.

5
👥 Build a team

Create more agents for different roles like reviewer or tester, and let them message each other.

6
📊 Monitor and adjust

View live progress, costs, and a graph of your team, pausing or resuming as needed.

✅ Enjoy autonomous help

Your AI team handles tasks independently while you stay in control with safety tools.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 13 to 14 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is AgentManager?

AgentManager is a TypeScript platform for running autonomous Claude Code agents at scale with human-on-the-loop oversight. It spins up isolated Claude Code CLI processes for tasks like coding, git ops, and MCP tool integrations (GitHub, Linear, Figma), while handling persistence across Cloud Run restarts via GCS. Developers get a Next.js UI for real-time streaming, agent graphs, cost dashboards, and API endpoints to spawn teams or send inter-agent messages.

Why is it gaining traction?

Unlike SDK wrappers, it runs real Claude Code sessions with built-in tools, auto-resuming state and delegating via a message bus—perfect for coordinated dev teams without prompt juggling. The 6-layer kill switch (process kills, token rotation, remote GCS halt) stems from a real rogue agent incident, plus cgroup memory limits and command blocklists add safety absent in basic agent runners. Cost tracking per model (Opus/Sonnet/Haiku) and cron scheduling hook users tired of runaway bills.

Who should use this?

Backend engineers orchestrating AI code reviews or PR monitors, AI researchers testing multi-agent workflows, or ops teams automating git/MCP tasks like Linear issue triage. Ideal for those hitting Claude Code limits in production, needing persistence and safeguards without rebuilding from scratch.

Verdict

Try it for Claude Code scaling—strong docs, demo video, and Terraform deploys make setup fast, but 13 stars and 1.0% credibility signal early-stage risks; test locally first before prod.

(187 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.